# Scott Clark — Full Site Corpus This file concatenates every page on https://scottclark.io as plain markdown for one-shot LLM ingestion. Built automatically at deploy time. Per-page sources are at https://scottclark.io/index.md, /cv.md, /talks.md, /projects.md, /publications.md, /press.md, /blog.md, and `/blog/.md`. ================================================================ # Scott Clark > I build startups that bring AI research into production at enterprise scale. Scott Clark is the co-founder and CEO of **[Distributional](https://distributional.com)**, building **[Talaria Scientific](https://talariasci.com)**, a multi-agent harness for computational science research. He was previously co-founder and CEO of **SigOpt**, a YC-and-a16z-backed Bayesian optimization platform acquired by **Intel** in 2020, where he then served as VP & GM of AI and HPC supercomputing through 2023. He holds a PhD in applied mathematics and an MS in computer science from Cornell University, where he was a Department of Energy Computational Science Graduate Fellow. ## At a glance **Researcher.** Published across multiple STEM fields. 1,200+ citations, h-index 16. ~20 granted US patents. PhD Applied Math and MS Computer Science from Cornell University; BS Mathematics, BS Physics, and BS Computational Physics from Oregon State University **Founder.** Two-time AI-startup founder, backed multiple times by a16z and Two Sigma. SigOpt (YC W15): raised $17M, acquired by Intel in 2020. Distributional: raised $30M, serving multiple Fortune 500 customers. **Operator.** Built and led teams from 2 to 200. From AI startups as a YC founder to the AI and HPC engineering organization at Intel as VP & GM. Frequent invited speaker at conferences and podcasts. ## Writing Recent essays (all: https://scottclark.io/blog; markdown index: https://scottclark.io/blog.md): - [How Better Evals Can Bring Abundance through Accelerated Scientific Discovery (talk recap)](https://scottclark.io/blog/better-evals-abundance.md) (August 31, 2026, Building Things): My five-minute lightning talk from the Agentic AI Summit at UC Berkeley: why evals are the hinge between AI and scientific abundance, how two failed products taught me where that hinge breaks, and what I'm building because of it. Video, slides, and a slide-by-slide walkthrough. - [Making New Mistakes Faster as a Second-Time Founder](https://scottclark.io/blog/make-all-new-mistakes-faster.md) (July 28, 2026, Startup Lessons): Lessons from a hard pivot: overfitting to my first company's successes, vision-market fit vs product-market fit, and making new mistakes faster. ## Experience ### 2023 — present · Distributional · Co-founder & CEO - Building [Talaria Scientific](https://talariasci.com): a multi-agent harness for computational science that turns a 10x researcher into a 100x researcher - Built AI testing and agent analytic tools before [pivoting in July 2026](https://distributional.com/blog/distributional-is-now-talaria) - Raised $30M from a16z, Two Sigma Ventures, SV Angel, and others ### 2020 — 2023 · Intel · VP & GM, AI/HPC Supercomputing - Led a team of 200 engineers and technologists worldwide, overseeing a $70M annual budget - Responsible for AI and HPC application quality on 35MW+ government datacenters - Promoted from Director to Sr. Director to VP, supporting multiple chip and datacenter launches ### 2014 — 2020 · SigOpt · Co-founder & CEO - Bayesian Optimization-as-a-Service product based on my PhD and open source work - Led a team of 25; published research at top conferences and closed several million in ARR - Acquired by Intel. Raised $17M from a16z, Two Sigma, YC (W15), others over 3 rounds ### 2012 — 2014 · Yelp · Software Engineer & Team Lead, Ad Targeting - Led MOE team (Metric Optimization Engine), 1.3k+ star open source repo based on my PhD - Developed novel location-based and bandit-optimized ad targeting algorithms - Created the Yelp Dataset Challenge, used by hundreds of thousands of students globally ## Education ### 2008 — 2012 · Cornell University · Ph.D. Applied Mathematics, M.S. Computer Science **Department of Energy Computational Science Graduate Fellow** (full four-year scholarship). Dissertation: "Parallel Machine Learning Algorithms in Bioinformatics and Global Optimization." Advisor: **Peter Frazier**. Committee: **Steve Strogatz**, **Bart Selman**. DOE practicums at Los Alamos National Laboratory and the Joint Genome Institute (LBNL). ### 2004 — 2008 · Oregon State University · B.Sc. Mathematics, B.Sc. Computational Physics, B.Sc. Physics **Triple bachelor's degrees** in four years, *magna cum laude*. Minors in Actuarial Sciences and Mathematical Sciences. Paradigms in Physics degree track. NSF REU summers at UC Davis (computational biophysics) and the Max Planck Institute Dresden (extreme-value statistics of chaotic quantum systems). ## Links - [GitHub](https://github.com/sc932) - [LinkedIn](https://www.linkedin.com/in/sc932) - [Scholar](https://scholar.google.com/citations?user=mwWbhAUAAAAJ) - [ORCID](https://orcid.org/0009-0007-0478-7129) - [@DrScottClark](https://twitter.com/DrScottClark) - [scott@scottclark.io](mailto:scott@scottclark.io) --- Source: https://scottclark.io/ ================================================================ # Scott Clark — Curriculum Vitae Long-form CV. The short version lives on the home page (https://scottclark.io/). Downloadable PDFs: [two-page resume](https://scottclark.io/resume/scott-clark-resume.pdf) and the longer [academic CV](https://scottclark.io/resume/scott-clark-cv.pdf). Full LaTeX source and a structured markdown profile are at https://github.com/sc932/resume. ## Experience ### 2023 — present · Distributional · Co-founder & CEO - Pivoted to **[Talaria Scientific](https://talariasci.com)** in 2026 (same corporation, same investors, new mission): a multi-agent harness for computational science — scientist-in-the-loop agents with HPC as the guardrail, in private beta. [The announcement](https://distributional.com/blog/distributional-is-now-talaria). - Analytics for AI agents — discovering behavioral signals in agent trace data for continuous AI reliability. **30-person team**, **$30M raised** (Seed Dec 2023 led by *Andreessen Horowitz*; Series A Oct 2024 led by *Two Sigma Ventures*). - Co-founders: Michael McCourt (CTO through ~2025; multi-paper SigOpt-era co-author), David Rosales (COO), Nick Payton (CRO). 11-person founding team sourced from *Bloomberg*, *Google*, *Meta*, *Intel*, *SigOpt*, *Slack*, *Stripe*, *Uber*, *Yelp*. - Product evolved 2023–2025 from pre-deployment AI-testing to production behavioral analytics for AI agents; sunset with the 2026 pivot to Talaria. - **US Patent 12,505,027** (2025, named inventor): anomaly detection in deployed AI applications. ### 2020 — 2023 · Intel · VP & GM, AI/HPC Supercomputing - Joined Intel through the acquisition of SigOpt (Oct 2020). Led a multi-disciplinary **~200-person organization** responsible for application-level AI and HPC software within Intel's Supercomputing Group. - **Title progression:** Director & GM of SigOpt (Nov 2020 – Mar 2021) → Senior Director & GM of SigOpt and AI Application Enablement (Mar 2021 – Sep 2022) → VP & GM of AI and HPC Supercomputing Application Level Engineering (Sep 2022 – Jun 2023). - Public face for Intel's **oneAPI** open-ecosystem thesis — *SC22 theCUBE* panel (Dallas, Nov 2022), MLconf 2021 Webinar, Ai4 2021, *Intel Conversations in the Cloud* Ep. 250. - Integrated SigOpt's IP into Intel's AI Analytics Toolkit; recognized as a leader in AI Software Optimization (*Kisaco Research Leadership Council*, 2021). ### 2014 — 2020 · SigOpt · Co-founder & CEO - Commercial Bayesian-optimization platform serving Fortune 500 enterprises in finance, trading, intelligence, and technology. **Acquired by Intel** (Oct 29, 2020). - **Raised $17M** across seed + Series A + Series A+ + strategic rounds: *Andreessen Horowitz* (led seed 2015 and A 2016); *Blumberg Capital* (lead, 2018 Series A+, with *Two Sigma Investments* co-investing, announced 2019); *Data Collective (DCVC)*, *SV Angel*, *Stanford*, *In-Q-Tel* strategic, *Y Combinator (W15)*, and notable angels. - Co-founders: Patrick Hayes (CTO 2014–2021), Eric Liu. - Grew from **Y Combinator W15** through Series A+ with customers in algorithmic trading, government/intelligence, enterprise AI, and consumer-tech. - Awards: *Gartner Cool Vendor in AI Core Technologies* (2017); *Barclays Open Innovation Challenge* winner (2017); *Kisaco Research KLC Leader* (2021). - Drove research culture and IP: **15 peer-reviewed papers** + **~20 granted US patents** as named inventor across Intelligent Optimization Platform, Multi-Criteria Optimization, and Multi-Solution Hyperparameter Tuning families. ### 2012 — 2014 · Yelp · Software Engineer & Team Lead, Ad Targeting - **Optimization:** Co-developed and led team for **MOE** (Metric Optimization Engine, github.com/Yelp/MOE) — first production Bayesian-optimization open-source package. Third-most-popular Yelp open-source release within 72 hours of launch; basis of SigOpt's founding. - **Ad targeting:** Implemented multi-armed bandit strategies for ad selection, sole targeting engineer on mobile ads rollout, developed location-based targeting algorithms. - **Director, Yelp Dataset Challenge:** created, implemented, and directed yelp.com/dataset_challenge; used by 100,000+ students worldwide since its inception. - **Leadership:** Founded Yelp's internal Applied Learning Group (bi-weekly all-engineering speaker series); MOE team lead; intern and new-hire mentor; 200+ technical interviews. - Featured in the *Wall Street Journal*, Aug 8 2014 (Elizabeth Dwoskin, "Big Data's High-Priests of Algorithms") and the Cornell ORIE alumni spotlight (2014). ### May — Aug 2011 · Bloomberg LP · Financial Software Development Intern - Developed end-to-end portfolio-analytics function in C++ and JavaScript (concept → back-end integration → GUI → PDF reporting). ## Education ### 2008 — 2012 · Cornell University · Ph.D. Applied Mathematics, M.S. Computer Science **Department of Energy Computational Science Graduate Fellow** (full four-year scholarship). Dissertation: "Parallel Machine Learning Algorithms in Bioinformatics and Global Optimization." Advisor: **Peter Frazier**. Committee: **Steve Strogatz**, **Bart Selman**. DOE practicums at Los Alamos National Laboratory and the Joint Genome Institute (LBNL). ### 2004 — 2008 · Oregon State University · B.Sc. Mathematics, B.Sc. Computational Physics, B.Sc. Physics **Triple bachelor's degrees** in four years, *magna cum laude*. Minors in Actuarial Sciences and Mathematical Sciences. Paradigms in Physics degree track. NSF REU summers at UC Davis (computational biophysics) and the Max Planck Institute Dresden (extreme-value statistics of chaotic quantum systems). ## Research Experience ### May — Aug 2010 · DOE Joint Genome Institute (LBNL) · DOE CSGF Practicum under Dr. Zhong Wang & Rob Egan - Built open-source genome-assembly validation framework (**ALE**, published *Bioinformatics* 2013, 208 cites). ### May — Aug 2009 · Los Alamos National Laboratory · DOE CSGF Practicum under Drs. Nick Hengartner & Joel Berendzen - Metagenomics / local sequence-alignment algorithms in Python, C, and CUDA (*Velvetrope*). This practicum pivoted my dissertation from computational fluid dynamics to bioinformatics. ### May — Aug 2007 · Max Planck Institute for the Physics of Complex Systems · NSF REU under Prof. Steven Tomsovic (WSU) - Extreme-value statistics of chaotic quantum systems in MATLAB and FORTRAN. ### May — Aug 2006 · University of California, Davis · NSF REU under Prof. Daniel Cox - Computational biophysics / protein folding in Java. Results published in *Prion* (2008). ## Awards & Honors ### 2016 · Forbes 30 Under 30 Enterprise Technology category. Recognized as co-founder & CEO of SigOpt. ### 2016 · Young Alumni Award Oregon State University, College of Science. ### 2010 · DOE CSGF Communicating Science Award (Honorable Mention) For the essay "Solving Genomic Jigsaws," *DEIXIS Magazine*. ### 2008 — 2012 · DOE Computational Science Graduate Fellowship Full four-year PhD scholarship (~$300,000), Department of Energy. Contract DE-FG02-97ER25308. ### 2008 · Cornell University Sage Fellowship $55,000, declined in favor of DOE CSGF. ### 2010 — 2012 · NERSC Production + Startup Allocations Principal Investigator, Cray XT4 (DOE Contract DE-AC02-05CH11231), 100,000 production hours plus startup renewals. ## Board & Advisory ### 2022 — 2025 · Oregon Museum of Science and Industry · Board of Trustees · Treasurer · Finance Committee Chair Executive Committee member; chaired the Finance Committee through OMSI's post-COVID financial recovery. ### 2019 — 2025 · Oregon State University, College of Science · Board of Advisors Advised the Dean's office on industry partnerships and research-to-product translation. ### 2018 — 2025 · Oregon State University, College of Science · Industry and Innovation Council Industry-side member supporting the College's translational research initiatives. ## Research Areas Bayesian optimization · Gaussian processes · Optimal learning · Multi-armed bandits · Hyperparameter tuning · AI reliability · AI testing & evaluation · Production ML observability · LLM and agent evaluation · Experiment design · A/B testing · Monte Carlo methods · Numerical analysis · High-performance computing · Parallel algorithms · Distributed systems · CUDA / GPU computing · Bioinformatics · Genome assembly · Computational biophysics ## Selected Publications ### 2020 · Parallel Bayesian Global Optimization of Expensive Functions *Operations Research*. Wang, Clark, Liu, Frazier. 263+ citations. Theoretical and practical algorithms for parallel evaluation in BO. ### 2013 · ALE: A Generic Assembly Likelihood Evaluation Framework *Bioinformatics*. Clark, Egan, Frazier, Wang. 208+ citations. Reference-free quality metric for genome assemblies. Output of the JGI 2010 DOE practicum. ### 2016 · Bayesian Optimization for Machine Learning: A Practical Guidebook *arXiv:1612.04858*. Dewancker, McCourt, Clark. 142+ citations. Widely-cited practitioner guide; the most-read public artifact from the SigOpt research program. ### 2025 · Anomaly Detection in Deployed Artificial Intelligence Applications *U.S. Patent 12,505,027*. [Distributional](https://distributional.com)'s first granted patent. McCourt, Bourassa-Denis, Laban, Kim, Dewancker, Cheng, Clark. Full publication list: https://scottclark.io/publications (1,200+ citations, h-index 16; Google Scholar: https://scholar.google.com/citations?user=mwWbhAUAAAAJ) ## Selected Talks & Podcasts ### Mar 2026 · AI Reliability for the Enterprise *SAIR Podcast* ### May 2025 · Behavioral Analytics for Production AI *AI + a16z Podcast (with Matt Bornstein)* ### Nov 2022 · oneAPI and the Future of Accelerated Computing *SC22 theCUBE Panel (with David Schmidt of Dell)* ### Mar 2019 · Tuning the Untunable *NVIDIA GTC Silicon Valley* ### Dec 2019 · From argmax f(x) to an International Business *Cornell CAM Notable Alumni Speaker Series* Full talks/podcasts list (40+ appearances since 2013): https://scottclark.io/talks --- Source: https://scottclark.io/cv ================================================================ # Talks & Podcasts — Scott Clark 40+ documented conference and podcast appearances since 2013. ## How Better Evals Can Bring Abundance through Accelerated Scientific Discovery **2026-08-02** · Agentic AI Summit 2026, UC Berkeley [Video](https://www.youtube.com/watch?v=1VO7hdEgSeo) · [Slides](/slides/better-evals-abundance.pdf) Five-minute lightning talk in the agent evaluation and benchmarks session: evals as the hinge between AI and scientific abundance, the grad-student-descent arc from SigOpt through two failed products, and the computational science harness now being built on those lessons. Recap with video and slide-by-slide walkthrough at /blog/better-evals-abundance. ## How Agents Can Lead to Abundance **2026-07-18** · AGI Summit 2026, San Francisco Thirty-minute talk on accelerating science with agentic AI: the augmentation-over-automation case, why validation pace sets the pace of science, and what changes when months of computational work collapse toward hours. ## AI as Science's Greatest Translator **2026-03-03** · SAIR Podcast (with Chuck Ng) [Video](https://www.youtube.com/watch?v=7qDx8PdljiA) Conversation with SAIR co-founder Chuck Ng on collaboration, observability, and AI's role as a translation layer for scientific work. ## AI Testing and Evaluation **2025-07-13** · GenAI Week Silicon Valley 2025 Speaker at GenAI Week Silicon Valley 2025 (Santa Clara Convention Center, July 13–17) on the AI observability landscape and the importance of behavioral signals in AI evaluation systems for enterprise reliability. ## Coffee with a Founder: Scott Clark, Distributional **2025-06-03** · BAM Podcast [Video](https://www.youtube.com/watch?v=qYcphOXSSEE) Founder-interview format conversation on building [Distributional](https://distributional.com), the enterprise AI reliability thesis, and the path from SigOpt through Intel to founding again. ## Building AI Systems You Can Trust **2025-05-23** · AI + a16z Podcast [Video](https://podcasts.apple.com/us/podcast/building-ai-systems-you-can-trust/id1740178076?i=1000709586075) Co-hosted conversation with Matt Bornstein (a16z partner) on enterprise AI reliability, agentic systems, and how Distributional approaches behavioral testing of production AI. ## Distributional Co-Founder & CEO Interview **2025-02-18** · CEO.com Podcast [Video](https://www.youtube.com/watch?v=I5btboBQIIk) Founder-CEO conversation on building Distributional, the enterprise AI testing thesis, and lessons from founding two ML companies. ## The Hidden Signal in Production AI Logs **2025-01-01** · Jason Liu Podcast [Video](https://www.youtube.com/watch?v=FKL918FgxAw) Long-form conversation with Jason Liu on production AI observability, what behavioral signals to extract from production logs, and the testing thesis behind Distributional. ## NYSE Distributional Spot **2024-01-01** · NYSE [Video](https://www.youtube.com/watch?v=RBCbinZ_UV0) Short NYSE-branded clip introducing Distributional and the enterprise AI testing thesis. ## Power Consumption and AI Are Key Priorities for Dell and Intel in the Supercomputing Space **2022-11-15** · SuperComputing 22 (theCUBE panel, Dallas) [Video](https://www.youtube.com/watch?v=-B-vQ0koA3Y) theCUBE panel at SuperComputing 22 in Dallas with David Schmidt of Dell Technologies. Discussion of Intel's oneAPI open-ecosystem strategy across heterogeneous hardware. Speaking as Intel's VP & GM of AI and HPC Supercomputing Application Level Engineering. ## Fireside Chat with Sam Charrington: Experimentation in ML **2022-05-18** · TWIML AI Podcast Fireside Chat [Video](https://www.youtube.com/watch?v=GGpImwvgezA) Wide-ranging fireside conversation with Sam Charrington on the state of experimentation in machine learning, drawing on the SigOpt era and post-acquisition Intel work. ## How to Optimize Your Models with Intelligent AI Experimentation **2021-10-12** · MLconf 2021 Webinar [Video](https://www.youtube.com/watch?v=xqp7v3qTo0U) MLconf-hosted webinar on intelligent AI experimentation for model optimization, presented as SigOpt GM during the Intel era. ## Experimenting with AI Optimizations **2021-07-22** · Intel Conversations in the Cloud (Episode 250) [Video](https://podcasts.apple.com/us/podcast/experimenting-with-ai-optimizations-citc-episode-250/id552020357?i=1000529697823) Intel-produced podcast conversation with host Jake Smith on intelligent AI experimentation and how SigOpt's platform fits into Intel's broader AI software stack post-acquisition. ## Building the Better, More Scalable Algorithms **2021-04-29** · IT Visionaries (Mission.org) [Video](https://www.youtube.com/watch?v=RY7CF1W4SKQ) Post-acquisition conversation on Mission.org's IT Visionaries podcast about building scalable optimization algorithms, leading SigOpt under Intel, and the broader landscape of intelligent experimentation. ## Intelligent AI Experimentation **2021-01-01** · Ai4 2021 [Video](https://www.youtube.com/watch?v=dYnTfVdPPAI) Talk at the Ai4 2021 conference on intelligent AI experimentation as a foundation for production model development, presented as SigOpt GM during the Intel era. ## Boost AI Experimentation to Design, Explore, and Optimize Your Models **2021-01-01** · SigOpt Summit 2021 (keynote) [Video](https://www.youtube.com/watch?v=1BDa42BOwKo) Keynote at the virtual SigOpt Summit 2021 on accelerating AI experimentation across design, exploration, and optimization workflows. Co-speakers included Subutai Ahmad of Numenta. ## Scott Clark of SigOpt **2020-06-22** · AI at Work Podcast (PJC), Season 2 Episode 2 [Video](https://podcasts.apple.com/us/podcast/s2-episode-2-scott-clark-of-sigopt/id1519875578?i=1000479142061) Conversation on SigOpt's history and the Y Combinator origin story, covering the path from Yelp's MOE project to building an optimization-as-a-service company. ## From argmax f(x) to an International Business **2019-12-06** · Cornell CAM Notable Alumni Speaker Series [Video](https://www.youtube.com/watch?v=EUXRJs4GVg4) 76-minute career-arc talk at Cornell's Center for Applied Mathematics on how applied mathematics research in Bayesian optimization led from a Cornell PhD through Yelp's MOE project to founding SigOpt as an international optimization-as-a-service business. ## Automated Model Tuning **2019-11-01** · TWIML AI Podcast (Episode 324) [Video](https://twimlai.com/network/scott-clark/) Deep dive into automated model tuning techniques, covering the state of the art in Bayesian optimization and how it applies to production machine learning workflows. ## Supporting Rapid Model Development at Two Sigma **2019-06-15** · TWIML AI Podcast (Episode 273) [Video](https://twimlai.com/network/scott-clark/) Discussion on how SigOpt's optimization platform supports rapid model development workflows at scale, with a case study from Two Sigma. ## Best Practices for Scaling Modeling Platforms **2019-04-17** · O'Reilly AI Conference, New York 2019 [Video](https://www.youtube.com/watch?v=wL9BhmsG0sE) Co-presented with Matt Greenwood, Chief Innovation Officer at Two Sigma. Lessons from scaling SigOpt's intelligent experimentation platform alongside a Two Sigma case study on running a modeling platform at quantitative-finance scale. ## Modeling at Scale in Systematic Trading **2019-04-01** · Quantitative Finance Conference [Video](https://www.youtube.com/watch?v=CBwA6FodNxM) 54-minute talk on running model-development platforms at scale across algorithmic trading firms, drawing on work with funds representing $300B AUM and a Two Sigma case study. ## Tuning the Un-Tunable **2019-03-18** · NVIDIA GTC Silicon Valley 2019 [Video](https://www.youtube.com/watch?v=G00fVTKbmZE) Strategies for optimizing deep learning models with long training cycles using Bayesian optimization. Presented at NVIDIA GTC Silicon Valley 2019. ## A Conversation with Scott Clark **2017-10-16** · Voices in AI with Byron Reese, Episode 12 [Video](https://podcasts.apple.com/us/podcast/episode-12-a-conversation-with-scott-clark/id1291540809?i=1000495733050) 56-minute long-form conversation with Byron Reese covering algorithms, transfer learning, the nature of human intelligence, and broader philosophical territory in AI. ## Bayesian Optimization for Hyperparameter Tuning **2017-08-01** · TWIML AI Podcast (Episode 50) [Video](https://twimlai.com/network/scott-clark/) Podcast discussion on Bayesian optimization approaches to hyperparameter tuning, the theory behind optimal experiment design, and practical applications in production ML systems. ## Bayesian Global Optimization **2017-05-01** · MLconf Seattle 2017 [Slides](https://www.slideshare.net/SessionsEvents/scott-clark-ceo-sigopt-at-mlconf-seattle-2017) Deep dive into Bayesian global optimization methods, covering theory, algorithms, and practical applications for machine learning hyperparameter tuning at scale. ## Tuning Machine Learning Algorithms **2017-02-12** · The AI in Business Podcast (Emerj) [Video](https://podcasts.apple.com/us/podcast/tuning-machine-learning-algorithms-with-scott-clark/id670771965?i=1000381100550) Earliest documented podcast appearance, on Emerj's AI in Business Podcast (then 'AI in Industry'). Discussion of Bayesian optimization for tuning machine learning algorithms in production. ## Using Bayesian Optimization to Tune Machine Learning Models **2016-11-01** · MLconf San Francisco 2016 [Slides](https://www.slideshare.net/SessionsEvents/scott-clark-cofounder-and-ceo-sigopt-at-mlconf-sf-2016) Talk on applying Bayesian optimization techniques to efficiently tune machine learning model hyperparameters, drawing on experience building SigOpt's optimization platform. ## Adaptive Sequential Experimentation Techniques for A/B Testing and Model Tuning **2015-05-01** · The Web Conference (WWW 2015) Presentation on adaptive sequential experimentation methods that improve upon traditional A/B testing by dynamically allocating resources to the most promising alternatives. ## Introducing the Metric Optimization Engine (MOE) **2014-11-01** · MLconf San Francisco 2014 [Slides](https://www.slideshare.net/SessionsEvents/scott-clark-software-engineer-yelp-at-mlconf-sf) Introduction of MOE, an open-source Bayesian optimization framework built at Yelp for optimizing real-world metrics through intelligent experimentation. ## CMU Silicon Valley TOCS Colloquium **2013-04-09** · Carnegie Mellon University Silicon Valley [Video](https://www.youtube.com/watch?v=bOQqOL2en9M) Earliest documented public talk, given at CMU Silicon Valley's Talks on Computer Science colloquium during the Yelp tenure. Predates the 2014 MLconf SF MOE talk. --- Source: https://scottclark.io/talks ================================================================ # Projects & Patents — Scott Clark ## Projects ### Distributional Founded in 2023, [Distributional](https://distributional.com) built AI testing and agent behavioral-analytics tools for Fortune 500 enterprises: automated testing and statistical analysis of LLM and agent trace data for continuous AI reliability. In July 2026 it made a [hard pivot](https://distributional.com/blog/distributional-is-now-talaria) to [Talaria Scientific](https://talariasci.com) (same company, same investors), refocusing on a multi-agent harness for computational science. Raised $30M from Andreessen Horowitz, Two Sigma Ventures, and others. [Site](https://distributional.com) *Tags:* AI Testing, Behavioral Analytics, LLM Observability, Enterprise, Reliability ### SigOpt An intelligent experimentation platform that accelerated and amplified the impact of modelers everywhere using Bayesian optimization to efficiently tune models and run experiments. Y Combinator W15. Raised approximately $17M from Andreessen Horowitz, Blumberg Capital with Two Sigma, In-Q-Tel, and others. Acquired by Intel in November 2020. [Site](https://sigopt.com) *Tags:* Bayesian Optimization, Machine Learning, Experimentation, SaaS ### ALE (Assembly Likelihood Evaluation) A generic assembly likelihood evaluation framework for assessing the accuracy of genome and metagenome assemblies. ALE provides a reference-free quality metric using a probabilistic model of read placement. Originated during a DOE Joint Genome Institute practicum and published in Bioinformatics with 200+ citations. [Site](https://github.com/sc932/ALE) · [Repo](https://github.com/sc932/ALE) *Tags:* Bioinformatics, Open Source, Genomics, C ### MOE (Metric Optimization Engine) A global, black-box Bayesian optimization engine for real-world metric optimization. Built at Yelp, MOE implements state-of-the-art algorithms from Bayesian Global Optimization using Gaussian Processes and provides REST, Python, and C++ interfaces. [Site](https://github.com/Yelp/MOE) · [Repo](https://github.com/Yelp/MOE) *Tags:* Bayesian Optimization, Open Source, Python, C++ ### Yelp Dataset Challenge Yelp's flagship academic data competition, providing real-world business, review, and user data to students and researchers worldwide. Scott created, implemented, and directed the program at Yelp; it ran for multiple rounds beginning in 2013, expanded internationally in 2014, and has been used by over 100,000 students. [Site](https://www.yelp.com/dataset) *Tags:* Data, Research, Yelp ### sc932/resume — LaTeX Resume Template An open-source LaTeX template for academic and industry resumes, with 500+ GitHub stars. The same repository also hosts Scott's longer academic CV and a structured markdown profile, all generated from a single source. [Site](https://github.com/sc932/resume) · [Repo](https://github.com/sc932/resume) *Tags:* LaTeX, Open Source, Template ## Patents (as named inventor) ~20 granted US patents across 4 SigOpt patent families (2019–2025) plus 1 Distributional patent (2025). ### Accelerated Tuning / Advanced Curtailment [US 11,157,812](https://patents.google.com/patent/US11157812) (2021), [US 11,704,567](https://patents.google.com/patent/US11704567) (2023), [US 12,159,209](https://patents.google.com/patent/US12159209) (2024), [US 12,373,699](https://patents.google.com/patent/US12373699) (2025), [US 12,450,479](https://patents.google.com/patent/US12450479) (2025) Inventors: Michael McCourt, Jackle-Spriggs, Ben-Cheng Hsu, Howey, Vance, Jim Blomo, Patrick Hayes, Scott C. Clark Systems and methods for accelerating hyperparameter tuning through early curtailment of unpromising trials in a machine learning optimization service. ### Anomaly Detection in Deployed Artificial Intelligence Applications [US 12,505,027](https://patents.google.com/patent/US12505027) (2025) Inventors: Michael McCourt, Rachel Bourassa-Denis, Kyle Laban, Olivia Kim, Ian Dewancker, Bolong Cheng, Scott C. Clark Methods and systems for detecting anomalous behavior in production AI applications using statistical analysis of behavioral signals from inference traces. ### Intelligent Optimization Platform [US 10,217,061](https://patents.google.com/patent/US10217061) (2019), [US 10,282,237](https://patents.google.com/patent/US10282237) (2019), [US 10,379,913](https://patents.google.com/patent/US10379913) (2019), [US 10,445,150](https://patents.google.com/patent/US10445150) (2019), [US 10,565,025](https://patents.google.com/patent/US10565025) (2020), [US 10,607,159](https://patents.google.com/patent/US10607159) (2020), [US 11,163,615](https://patents.google.com/patent/US11163615) (2021), [US 11,301,781](https://patents.google.com/patent/US11301781) (2022), [US 11,709,719](https://patents.google.com/patent/US11709719) (2023), [US 12,141,667](https://patents.google.com/patent/US12141667) (2024), [US 12,236,287](https://patents.google.com/patent/US12236287) (2025) Inventors: Patrick Hayes, Michael McCourt, Alexandra Johnson, George Ke, Scott C. Clark Systems and methods for an intelligent experimentation and optimization platform that efficiently tunes parameters of expensive black-box models. ### Multi-Solution Hyperparameter Tuning [US 11,270,217](https://patents.google.com/patent/US11270217) (2022), [US 11,966,860](https://patents.google.com/patent/US11966860) (2024) Inventors: Kelvin Tee, Michael McCourt, Patrick Hayes, Scott C. Clark Systems and methods for generating multiple solution sets during hyperparameter tuning to support diverse model selection criteria. ### Multiple Tuning Criteria Optimization [US 10,528,891](https://patents.google.com/patent/US10528891) (2020), [US 10,558,934](https://patents.google.com/patent/US10558934) (2020), [US 10,621,514](https://patents.google.com/patent/US10621514) (2020), [US 10,740,695](https://patents.google.com/patent/US10740695) (2020), [US 11,699,098](https://patents.google.com/patent/US11699098) (2023), [US 12,033,036](https://patents.google.com/patent/US12033036) (2024) Inventors: Bolong Cheng, Olivia Kim, Michael McCourt, Patrick Hayes, Scott C. Clark, Jackle-Spriggs, Vance, Ben-Cheng Hsu, Frey Systems and methods for hyperparameter optimization across multiple competing objectives or constraints simultaneously. --- Source: https://scottclark.io/projects ================================================================ # Publications — Scott Clark 1,200+ citations · h-index 16 · Google Scholar: https://scholar.google.com/citations?user=mwWbhAUAAAAJ ### Parallel Bayesian Global Optimization of Expensive Functions Jialei Wang, **Scott C. Clark**, Eric Liu, Peter I. Frazier. *Operations Research*, 2020. [Link](https://pubsonline.informs.org/doi/10.1287/opre.2019.1966) DOI: 10.1287/opre.2019.1966 We develop parallel Bayesian global optimization methods for expensive black-box functions, providing both theoretical analysis and practical algorithms that enable efficient optimization across multiple parallel evaluations. ### Bayesian Optimization for Machine Learning: A Practical Guidebook Ian Dewancker, Michael McCourt, **Scott C. Clark**. *arXiv preprint*, 2016. [Link](https://arxiv.org/abs/1612.04858) We present a practical guide to Bayesian optimization for machine learning practitioners, covering the core concepts, common pitfalls, and best practices for hyperparameter tuning and model selection. ### Evaluation System for a Bayesian Optimization Service Ian Dewancker, Michael McCourt, **Scott C. Clark**, Patrick Hayes, Alexandra Johnson, George Ke. *arXiv preprint*, 2016. [Link](https://arxiv.org/abs/1605.06170) We describe the evaluation system used to benchmark and validate a production Bayesian optimization service, covering metrics, test functions, and evaluation methodologies. ### A Strategy for Ranking Optimization Methods using Multiple Criteria Ian Dewancker, Michael McCourt, **Scott C. Clark**, Patrick Hayes, Alexandra Johnson, George Ke. *AutoML Workshop at ICML 2016 (PMLR Vol. 64)*, 2016. [Link](https://proceedings.mlr.press/v64/dewancker_strategy_2016.html) We propose a multi-criteria strategy for ranking optimization methods, enabling principled comparison across diverse benchmark problems and performance metrics. ### A Stratified Analysis of Bayesian Optimization Methods Ian Dewancker, Michael McCourt, **Scott C. Clark**, Patrick Hayes, Alexandra Johnson, George Ke. *arXiv preprint*, 2016. [Link](https://arxiv.org/abs/1603.09441) We present a stratified analysis of Bayesian optimization methods, comparing their performance across different problem types and dimensionalities to provide guidance for practitioners. ### Adaptive Sequential Experimentation Techniques for A/B Testing and Model Tuning **Scott C. Clark**. *The Web Conference (WWW 2015)*, 2015. DOI: 10.1145/2740908.2743063 We present adaptive sequential experimentation techniques that improve upon traditional A/B testing by dynamically allocating resources to the most promising alternatives. ### ALE: A Generic Assembly Likelihood Evaluation Framework for Assessing the Accuracy of Genome and Metagenome Assemblies **Scott C. Clark**, Rob Egan, Peter I. Frazier, Zhong Wang. *Bioinformatics*, 2013. [Link](https://academic.oup.com/bioinformatics/article/29/4/435/199222) DOI: 10.1093/bioinformatics/bts723 We present ALE, a generic assembly likelihood evaluation framework that assesses the accuracy of genome and metagenome assemblies using a probabilistic model of read placement, providing a reference-free quality metric. ### Parallel Machine Learning Algorithms in Bioinformatics and Global Optimization **Scott C. Clark**. *PhD Dissertation, Cornell University*, 2012. This thesis develops parallel machine learning algorithms for two domains: bioinformatics (genome assembly evaluation) and global optimization (Bayesian optimization with parallel evaluations), with applications to real-world computational problems. ### Solving Genomic Jigsaws **Scott C. Clark**. *DEIXIS Magazine, DOE CSGF*, 2010. [Link](https://www.krellinst.org/csgf/profile/clark2008) A feature article in the DOE CSGF DEIXIS magazine describing computational approaches to genome assembly and the development of tools for evaluating assembly accuracy. ### Left-Handed Beta Helix Models for Mammalian Prion Fibrils K. Kunes, **Scott C. Clark**, Daniel L. Cox, Rajiv R. P. Singh. *Prion 2(2):81–90*, 2008. [Link](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2634523/) DOI: 10.4161/pri.2.2.7059 Statistical analysis of left-handed beta helix structural models for mammalian prion protein fibrils, applying computational biophysics methods to study protein misfolding. Output of an NSF REU at UC Davis under Prof. Daniel Cox. --- Source: https://scottclark.io/publications ================================================================ # Press — Scott Clark ## Meet Scott Clark, an OSU Science Alum Who Built a $30M AI Startup **2025-05-14** · *Oregon State University College of Science* · [Link](https://science.oregonstate.edu/impact/2025/05/meet-scott-clark-an-osu-science-alum-who-built-a-30m-ai-startup) > Clark's latest venture, Distributional Inc., is focused on a fast-growing challenge in today's AI landscape: reliability. It's already raised $30 million and grown to a team of 30. ## Distributional: Interview With Co-Founder & CEO Scott Clark About The Enterprise AI Testing Company **2025-03-31** · *Pulse 2.0* · [Link](https://pulse2.com/distributional-profile-scott-clark-interview/) > Our platform is designed to bridge the gap in AI testing, providing enterprise teams with the tools they need to proactively detect and address risks throughout the AI lifecycle. ## Distributional Raises $19M to Automate AI Model and App Testing **2024-10-08** · *TechCrunch* · [Link](https://techcrunch.com/2024/10/08/distributional-raises-19m-to-automate-ai-model-and-app-testing/) > Distributional, an AI testing platform, has secured $19 million in Series A funding led by Two Sigma Ventures. The company was established by Scott Clark, who previously served as Intel's general manager of AI software. ## Distributional Raises $11M Seed Round Led by Andreessen Horowitz to Make AI Safe, Secure and Reliable **2023-12-14** · *BusinessWire* · [Link](https://www.businesswire.com/news/home/20231214878185/en/Distributional-Raises-11M-Seed-Round-led-by-Andreessen-Horowitz-to-Make-AI-Safe-Secure-and-Reliable) > Distributional secured an $11M seed round led by Andreessen Horowitz with participation from Operator Stack, Point72 Ventures, SV Angel, Two Sigma, and 40+ angel investors. ## Power Consumption and AI Are Key Priorities for Dell and Intel in the Supercomputing Space **2022-11-18** · *SiliconANGLE / theCUBE* · [Link](https://siliconangle.com/2022/11/18/power-consumption-and-ai-are-key-priorities-for-dell-and-intel-in-the-supercomputing-space-sc22/) > Coverage of the SC22 theCUBE panel with Scott Clark (then VP & GM at Intel) and David Schmidt (Dell Technologies) on Intel's oneAPI open-ecosystem strategy and the future of accelerated computing for AI. ## Meet a Science Grad: Scott Clark **2020-12-18** · *Oregon State University College of Science* · [Link](https://science.oregonstate.edu/impact/2020/12/meet-a-science-grad-scott-clark) > Clark credits OSU's strong science faculty and innovative programs with laying the foundation for his Ph.D., which he turned into a startup. ## Intel to Acquire SigOpt to Scale AI Productivity and Performance **2020-10-29** · *Intel Corporation* · [Link](https://www.intc.com/news-events/press-releases/detail/1425/intel-to-acquire-sigopt-to-scale-ai-productivity-and) > By combining our AI optimization software with Intel's decades-long leadership in AI computing and machine learning performance, we will be able to unlock new capabilities for data scientists. ## Intel Acquires SigOpt, a Specialist in Modeling Optimization, to Boost Its AI Business **2020-10-29** · *TechCrunch* · [Link](https://techcrunch.com/2020/10/29/intel-acquires-sigopt-a-specialist-in-modeling-optimization-to-boost-its-ai-business/) > Intel announced an acquisition of SigOpt, a company specializing in modeling optimization technology. Co-founders Scott Clark and Patrick Hayes and their team joined Intel. ## Young Alumni Award Winner Makes Forbes' 30 Under 30 List **2016-11-23** · *Oregon State University College of Science* · [Link](https://science.oregonstate.edu/IMPACT/2016/11/young-science-alumni-award-winner-featured-forbes-30-30) > Scott Clark, a physics and mathematics alumnus from Oregon State University (Class of 2008), received the 2016 College of Science Young Alumni Award and was recognized among Forbes' 30 Under 30. ## Cornellians Named to Forbes 30 Under 30 List **2016-01-15** · *Cornell Chronicle* · [Link](https://news.cornell.edu/stories/2016/01/cornellians-named-forbes-30-under-30-list) > Scott Clark, who earned both a Master's degree and Ph.D. from Cornell, was recognized on the Forbes 30 Under 30 list in the Enterprise Tech category as co-founder and CEO of SigOpt. ## Using MOE to Optimize an A/B Testing Experiment Framework **2014-10-15** · *Yelp Engineering Blog* · [Link](https://engineeringblog.yelp.com/2014/10/using-moe-the-metric-optimization-engine-to-optimize-an-ab-testing-experiment-framework.html) > Follow-up on applying the Metric Optimization Engine to optimize A/B testing frameworks, demonstrating practical improvements in experiment efficiency at Yelp. ## Big Data's High-Priests of Algorithms **2014-08-08** · *Wall Street Journal* · [Link](https://www.wsj.com/articles/academic-researchers-find-lucrative-work-as-big-data-scientists-1407543088) > Academia is slow and only a few people see your work. At Yelp, I can be pushing out experiments that affect hundreds of millions of people. When I make a small change to the Yelp website, I have a bigger impact. ## Introducing MOE: A New Open Source Machine Learning Service for Optimal Experiment Design **2014-07-24** · *Yelp Engineering Blog* · [Link](https://engineeringblog.yelp.com/2014/07/introducing-moe-metric-optimization-engine-a-new-open-source-machine-learning-service-for-optimal-ex.html) > MOE is an open-source machine learning tool for solving black-box optimization problems using Bayesian Global Optimization algorithms and Gaussian Processes. ## Scott Clark Ph.D. '12 Brings His Work with ORIE Professor Frazier to Yelp and the Open Source World **2014-06-01** · *Cornell ORIE* · [Link](https://www.orie.cornell.edu/spotlights/scott-clark-phd-12-brings-his-work-orie-professor-frazier-yelp-and-open-source-world) > As one of the first Ph.D.'s hired by Yelp, Scott Clark built a black-box optimization system, Metrics Optimization Engine (MOE), developed in collaboration with ORIE Professor Peter Frazier. ## DEIXIS Magazine Feature: DOE CSGF Fellow Profile **2011-09-01** · *DEIXIS Magazine (DOE Computational Science Graduate Fellowship)* · [Link](https://www.krellinst.org/doecsgf/docs/deixis/deixis2011.pdf) > Substantive profile of Scott as a DOE Computational Science Graduate Fellow, covering the dissertation pivot from computational fluid dynamics to metagenomics during the 2009 Los Alamos National Laboratory practicum. --- Source: https://scottclark.io/press ================================================================ # Writing — Scott Clark > Essays from Scott Clark on startup lessons and building things. Each entry links the post's plain-markdown twin; HTML versions live at `https://scottclark.io/blog/`. Full-content RSS: https://scottclark.io/rss.xml. - [How Better Evals Can Bring Abundance through Accelerated Scientific Discovery (talk recap)](https://scottclark.io/blog/better-evals-abundance.md) (August 31, 2026): My five-minute lightning talk from the Agentic AI Summit at UC Berkeley: why evals are the hinge between AI and scientific abundance, how two failed products taught me where that hinge breaks, and what I'm building because of it. Video, slides, and a slide-by-slide walkthrough. - [Making New Mistakes Faster as a Second-Time Founder](https://scottclark.io/blog/make-all-new-mistakes-faster.md) (July 28, 2026): Lessons from a hard pivot: overfitting to my first company's successes, vision-market fit vs product-market fit, and making new mistakes faster. - [Building this blog: Part II](https://scottclark.io/blog/building-this-blog-part-ii.md) (July 28, 2026): How agents turned a decade-old blog backlog into a live site built for two readers: people, and the AI agents they send to read about you. --- Source: https://scottclark.io/blog (Scott Clark) ================================================================ # How Better Evals Can Bring Abundance through Accelerated Scientific Discovery (talk recap) > My five-minute lightning talk from the Agentic AI Summit at UC Berkeley: why evals are the hinge between AI and scientific abundance, how two failed products taught me where that hinge breaks, and what I'm building because of it. Video, slides, and a slide-by-slide walkthrough. Author: Scott Clark Published: 2026-08-31 Pillar: Building Things Canonical: https://scottclark.io/blog/better-evals-abundance --- On August 2, 2026, I gave a five-minute lightning talk at the Agentic AI Summit at UC Berkeley. The argument: better evals are the hinge between AI and real scientific abundance. I have spent 20 years working on versions of this problem, and [Talaria Scientific](https://talariasci.com) is the culmination of that work. The video is below; after it, the summary, then the whole talk slide by slide. [Watch on YouTube: How Better Evals Can Bring Abundance through Accelerated Scientific Discovery (Agentic AI Summit 2026)](https://www.youtube.com/watch?v=1VO7hdEgSeo) Prefer to page through the deck itself? [Download the slides (PDF)](https://scottclark.io/slides/better-evals-abundance.pdf). Want to learn more about Talaria Scientific? Read [why I am building Talaria](https://talariasci.com/blog/why-im-building-talaria), [how it works](https://talariasci.com/blog/the-talaria-architecture), and [what this unlocks](https://talariasci.com/blog/the-right-model-for-the-job). ## The summary Abundance through AI, in my mind, means new scientific discoveries: the kind that leads to new materials and more efficient energy, and from there to advances in medicine and aerospace. It means living in a sci-fi future. The real promise of AI is improving quality of life (not just replacing white-collar jobs). And I believe that for the first time, we can meaningfully accelerate science with AI. In grad school, the final step of any project was always parameter tuning, whether I was working on finite elements, protein folding, quantum mechanics, or metagenome assembly. We jokingly called it grad student descent: the least tenured person in the group, up late turning knobs in a high dimensional space, chasing a slightly better result. Solving that problem more generally became my thesis, and then my first company. SigOpt was Bayesian optimization as a service: given an eval, find the parameters that make it go up. Over seven years we tuned recommender systems at Netflix, fraud models at Amex, early RL systems at OpenAI, and models at hedge funds collectively managing about a trillion dollars, plus hundreds of academics who used our free program. In 2020 I sold the company to Intel. Somewhere in there I realized I was only solving half the problem, and I said it on stage the way I have said it for a decade: "The best thing about a black box optimizer is that it will optimize any eval you tell it to. The worst part about a black box optimizer is that it will optimize exactly the eval you tell it to." When only 1% of transactions are fraud, you can build a very accurate fraud detector by predicting that nothing is fraud. That gap between the eval you wrote and the outcome you care about is a big part of why we are still a long way from typing `/goal solve stable fusion containment` into anything. I started [Distributional](https://distributional.com) in 2023 to attack that gap head-on as an AI testing company, with high dimensional statistical tests for chaotic and non-stationary AI systems. It failed: people didn't know what to test, didn't have the data before deployment, and shipped to production anyway. We pivoted to post-production agent analytics, finding [the evals you should have written](https://distributional.com/blog/distributional-is-now-talaria). That failed too: a good idea, but more of a feature than a product. I still believe both approaches are obvious and inevitable; we just never crossed from [vision-market fit to product-market fit](https://scottclark.io/blog/make-all-new-mistakes-faster). So now I am doing what you should do when you have a problem: solving the version I care about most. Talaria Scientific ([Distributional's next chapter](https://distributional.com/blog/distributional-is-now-talaria)) is a multi-agent, multi-foundation-model harness for computational science research, built around domain-specific evals and guardrails rooted in real physics and math. The point is to spin the what-if flywheel faster: work that took me six months in grad school, done in six hours. It is in private beta now, it will be open source at NeurIPS 2026, and it will always be free for open science. ## The talk, slide by slide Each slide below carries what I said over it, lightly cleaned up from the livestream transcript. ![Title slide: How Better Evals Can Bring Abundance through Accelerated Scientific Discovery. Scott Clark, Founder and CEO, with the Distributional and Talaria Scientific logos.](https://scottclark.io/images/blog/better-evals-abundance/slide-01.png) My name is Scott Clark. I'm co-founder and CEO of Distributional, and I'm building Talaria Scientific. This talk is about how you get better abundance. ### What abundance with AI means ![Slide: what abundance with AI means, a four-step build from new scientific discoveries to living in a scifi future.](https://scottclark.io/images/blog/better-evals-abundance/slide-02.png) In my mind it's better scientific discoveries, which lead to new materials and more efficient energy, which lead to better medicine and more efficient ways to travel. It basically means living in a sci-fi future. ### The real promise of abundance through AI ![Slide: improving quality of life, not just replacing repetitive work.](https://scottclark.io/images/blog/better-evals-abundance/slide-03.png) It's less about how we get rid of more white-collar jobs, and more about how we improve people's quality of life. ### The pieces are finally falling into place ![Slide: we can now meaningfully accelerate science with AI.](https://scottclark.io/images/blog/better-evals-abundance/slide-04.png) I believe that for the first time, we are able to meaningfully accelerate scientific discovery with AI. The last few months have unlocked quite a bit of capability. ### A culmination of my life's work ![Slide: I have been trying to solve this problem for the last 20 years of my career, with varying degrees of success.](https://scottclark.io/images/blog/better-evals-abundance/slide-05.png) This has been a passion of mine for a long time. I've been trying to solve this problem for the last 20 years of my career, with varying degrees of success. ### Grad school: solving the same problem over and over ![Slide: grad school timeline across Oregon State, UC Davis, Max Planck, Cornell, Los Alamos, and Berkeley Lab. Whatever the field, the final step was always parameter tuning.](https://scottclark.io/images/blog/better-evals-abundance/slide-06.png) Every group I worked with, from protein folding to quantum mechanics simulation to metagenome assembly, ended with the same problem. We'd build something great, and then we'd need to tune it: knobs, levers, hyperparameters, whatever you want to call them. If you could make the benchmarks slightly better, you got a better paper. We jokingly called this grad student descent, because it was usually the grad students sitting up late, tuning knobs in a high dimensional space. People applied smart techniques to it (simulated annealing, genetic algorithms, local methods), but I fell in love with Bayesian optimization, and it became the core of my PhD thesis: methods for efficient optimization, given an eval. ### SigOpt: optimize everything ![Slide: SigOpt, Bayesian optimization as a service, with the founding, building, and scaling timeline.](https://scottclark.io/images/blog/better-evals-abundance/slide-07.png) That thesis became SigOpt, Bayesian optimization as a service, started in 2014. Over seven years we worked with Netflix tuning recommender systems, Amex tuning fraud systems, OpenAI tuning their early RL systems back when they were a nonprofit lab, hedge funds managing about a trillion dollars, and several hundred academics who used our free program on everything from materials design to drug discovery. After I sold the company to Intel in 2020, we tuned everything from chip design to benchmaxxing MLPerf, back when that was the number everyone was overfitting to. ### Learning I was only solving half the problem ![Slide: the optimizer worked, except when it didn't. It overfit, or optimized the wrong thing.](https://scottclark.io/images/blog/better-evals-abundance/slide-08.png) It worked really well, except when it didn't. It would always optimize exactly what you gave it. Customers would come back and say: you made that number go up, but some other number went down. ### The problem with optimizers ![Slide: the best thing about a black box optimizer is that it will optimize any eval you tell it to. The worst part is that it will optimize exactly the eval you tell it to.](https://scottclark.io/images/blog/better-evals-abundance/slide-09.png) The best thing about a black box optimizer is that it will optimize any eval you tell it to. The worst part about a black box optimizer is that it will optimize exactly the eval you tell it to. You can build a very accurate fraud detector, when only 1% of your transactions are fraud, by predicting that nothing is fraud. ### We are still a long way from `/goal solve` ![Slide: we are still a long way from typing goal solve stable fusion containment.](https://scottclark.io/images/blog/better-evals-abundance/slide-10.png) We're a long way from saying: goal, solve stable fusion containment. Or solve cancer. And even if the model came back with something, how could we trust it? That's the other side of the coin. You need to optimize these systems, and you need to trust them. ### Have a problem? Start a startup! ![Slide: Distributional attempt one, pre-production AI testing, 2023 to 2025: high dimensional Bayesian statistical tests for chaotic and non-stationary AI systems and agents.](https://scottclark.io/images/blog/better-evals-abundance/slide-11.png) That's what I set out to do with Distributional in 2023. I attacked it in the most complex mathematical way possible: high dimensional statistical tests for chaotic and non-stationary AI systems. ![Slide: the same testing slide grayed out with It failed stamped over it, and the three reasons: people didn't know what to test or have the data before deployment, they couldn't understand tests well enough to act, and they shipped to prod anyway.](https://scottclark.io/images/blog/better-evals-abundance/slide-12.png) It failed. Nobody likes tests. People didn't know what to test, they didn't have the data before deployment, and everybody was YOLOing models into production anyway. ### Have a problem? Pivot! ![Slide: Distributional attempt two on the timeline strip, post-production agent analytics, 2025 to 2026: analytics on agent traces to find the evals you aren't monitoring for.](https://scottclark.io/images/blog/better-evals-abundance/slide-13.png) So I pivoted the company. This is what you do: you're failing, pivot. Analytics. Catch the evals that are sneaking through your system, find the patterns in behavior that say, "this is the unknown unknown," the eval you should have written. ![Slide: the pivot slide grayed out with It failed stamped over it. A feature, not a product, let alone a business or a startup.](https://scottclark.io/images/blog/better-evals-abundance/slide-14.png) This also failed. It's a really good idea, but it's more of a feature than a product. Let alone a business, or a startup. ### Have a problem? Solve the version you care about ![Slide: the timeline strip gains its third column, Computational Science Harness, 2026 and beyond: a multi-agent, multi-FM harness for computational science research acceleration, using domain-specific evals and guardrails rooted in physics and math.](https://scottclark.io/images/blog/better-evals-abundance/slide-15.png) So now I'm applying everything I've done in the last 20 years to attack the scientific problem: how do we accelerate science? Talaria is a multi-agent, multi-foundation-model harness for computational science research, focused on what actually matters: bespoke evals that know physics, that know the scientific research process, that know how to do hypothesis validation. ### From idea to trusted result ![Slide: from idea to trusted result, where the week goes. Two rows compare a computational scientist's week today, dominated by the validation schlep done by hand, with the same week in the harness, where data wrangling, environments, baselines, sweeps, and checking run with provenance on every run.](https://scottclark.io/images/blog/better-evals-abundance/slide-16.png) With that in place, you can spin the flywheel of a what-if machine faster and faster. Work that used to take me six months in grad school takes six hours instead. ### Thank you ![Slide: thank you, with QR codes for talariasci.com and scottclark.io. Researching computational physics or applied math? Building world models? Private beta signups, open source launch at NeurIPS 2026, always free for open science.](https://scottclark.io/images/blog/better-evals-abundance/slide-17.png) We're building this in the open: private beta signups are open at talariasci.com now, the open source launch lands at NeurIPS 2026, and it will always be free for open science. I would love to chat if you're building anything computational science related. ## If you're building in this space If you're working on world models, materials, aerospace, computational physics, or optimization, I would love to hear from you: [talariasci.com](https://talariasci.com). The longer version of the Distributional story is in [Making New Mistakes Faster as a Second-Time Founder](https://scottclark.io/blog/make-all-new-mistakes-faster), and the full case for what we're building is in the [Talaria manifesto](https://talariasci.com/blog/why-im-building-talaria). *Recorded at the [Agentic AI Summit 2026](https://rdi.berkeley.edu/events/agentic-ai-summit-2026) (UC Berkeley RDI), Compass Stage, August 2, 2026. Thanks to the summit crew for the livestream; the [full session stream](https://www.youtube.com/watch?v=1UrriPJRSPU&t=9785s) has my talk at 2:43:05.* --- Follow me at [@DrScottClark](https://twitter.com/DrScottClark) to see new posts. Source: https://scottclark.io/blog/better-evals-abundance (Scott Clark) ================================================================ # Making New Mistakes Faster as a Second-Time Founder > Lessons from a hard pivot: overfitting to my first company's successes, vision-market fit vs product-market fit, and making new mistakes faster. Author: Scott Clark Published: 2026-07-28 Pillar: Startup Lessons Canonical: https://scottclark.io/blog/make-all-new-mistakes-faster --- When I started [Distributional](https://distributional.com) in 2023 I thought I had it all figured out. I had sold my first company, SigOpt, to Intel in 2020 and had spent the previous 3 years as an executive learning from the inside of one of the most iconic tech companies of all time. I had built SigOpt from my PhD thesis in 2014 and ran it for 7 years before the acquisition. I made many, many mistakes along the way as I learned company building, management, and enterprise sales mostly by trial and error. After reflecting on a decade of mistakes and learnings I was ready to embark on a new adventure with all of my experience earned through blood, sweat, and tears. I thought I was going to be able to accelerate through the first few stages of company building because I would be able to avoid old mistakes, repeat what worked, and make all new mistakes faster. What I didn't account for was that the entire market and world was fundamentally shifting, so many old lessons didn't apply directly. I also overfit to my previous successes, which turned them into mistakes, [forcing us to pivot multiple times along the way](https://distributional.com/blog/distributional-is-now-talaria). I had to learn the hardest lesson again: startups are hard, there are no shortcuts to product-market fit, and you need to do it the hard way for it to have any chance of working. There is nothing like the free market to humble you, but the upside is that there is still so much to learn. Here is what I did right, wrong, and what I will do differently [as I keep building](https://talariasci.com/blog/why-im-building-talaria). ## What worked the first time On paper, SigOpt was a great run. We went through Y Combinator in Winter 2015, raised from top investors like a16z and SV Angel, took the Bayesian optimization tooling I'd built in grad school and open sourced at Yelp and turned it into an enterprise product, grew the team, stood up a sales motion, and sold the company to Intel in 2020. Underneath those highlights was a first-time founder figuring it out as we went, often via trial and error. The thing that saved us over and over again was the incredible team and culture I was able to build with my co-founder. But the big timing calls mostly worked: when we hired, when we scaled, when we leaned into go-to-market, when we sold. I took some of our luck and attributed it to skill, forgetting some of the misfires along the way. Hiring ahead of product-market fit had worked, because the ML-infrastructure market grew into the capacity we built. Standing up sales early had worked, because buyers were already forming budgets around our category. Eight years later, none of that felt like a playbook I was choosing to re-run. It felt like knowing what I was doing. ## Success is a worse teacher than failure Mistakes are often obvious immediately or at least in hindsight. You miss the quarter, you lose the deal, you watch a launch land flat, and the lesson arrives with a timestamp and a bruised ego. Failure hands you a signal to reinforcement learn from. Success gives you almost no information about why it worked. Was it the play? The timing? The market regime? Luck? A single evaluation on one company outcome can't separate those variables. Even SigOpt's ending made the point: we were mid-raise, term sheets all but verbally agreed, when the news broke in a fundraising meeting with one of our current investors that the NBA had suspended its season due to COVID. The pandemic changed the world, the round evaporated, and we sold the company later that year. A good outcome by any startup standard, and the variable that timed it appears in nobody's playbook. A founding career is a training set with a few samples per company, and I fit a confident model to mine. Overfitting is roughly the first danger you learn to fear in getting a PhD, and I did it anyway, with my own resume as the training data. It gets worse, because the environment isn't stationary. The AI market between 2015 and 2023 didn't drift, it monotonically grew, and by the end it was jumping several times a year: categories forming and dissolving in months, buyers rethinking budgets mid-cycle. My old plays weren't wrong in some universal sense; they were fit to a distribution that no longer existed. Success can often tilt your policy from exploration all the way toward exploitation. You've found high ground before, so you march straight at where it was. When the terrain moves, the value of exploration goes back up, and your own experience is the voice insisting you don't need it. And when building a startup the terrain is always moving, sometimes catastrophically. ## What re-running the playbook looked like I'll always remember what one of our board members said after we went through our first pivot and reduction in force: "Every great company needs to figure out product, sales, operations. But you need to do it sequentially and in that order." The first few years of Distributional were about me trying to do it all at once. I wanted to skip past the parts I had already done and get to new ground as fast as possible. In doing so I starved the part that mattered the most: building something people actually want. By trying to scale up sales and operations I thought I was "spring loading" the business for the second we found product-market fit. But in reality I starved our ability to focus on product or iterate as fast as the changing market demanded. I hired and scaled ahead of product-market fit, because last time that timing had been right. This mistake compounds quietly, step by step. You hire ahead of fit, and the new team needs a roadmap to build and sell. A roadmap demands conviction you don't actually have yet, so you commit to your best current guess. But the team also grows so big that you need to operationalize things that should be fluid. Once the org is executing on that process, changing it costs a reorg, so each piece of contrary evidence gets discounted a little. Momentum and process start being about the machine instead of the market. From inside, it looks like execution, but it really is starving innovation and speed. The result is all of the stress of a growing startup, with none of the results. ## Naming the quadrant We had vision-market fit more than once. Smart, senior people in our ideal customer profile agreed the problem was real and coming: AI systems need testing you can trust. People were excited to try the product when we were pitching. When we repositioned the product, the excitement was even greater. People agreed it was obvious and inevitable and we had a unique and valuable take on the problem. Every signal a founder wants to see, except the one that pays. Getting close is worse than not being close at all, because close is exactly the feeling that keeps you funding the push. It took me too long to treat that pattern as a distinct false positive with a name, and that turned out to be the lesson. *Vision-market fit* is resonance: people with real budgets and real problems agree that what you see coming is coming, and interest shows up wherever you push. The tell is what never happens next, if anything. The interest stays excitement instead of execution: nothing gets deployed because it wasn't a real pain. I remember in the early days of SigOpt, when we were selling hyperparameter optimization as a service to firms just learning about ML, we would sometimes get the response, "I can't wait until we have this problem, that will mean we have a system worth optimizing." In SigOpt we just had to go up market to find product-market fit, but with Distributional the whole field and concept we were pushing was so nascent that it meant there was no up market to chase, even the most advanced firms of the ML era were still figuring it out as they went. *Product-market fit* is pull: buyers you didn't call show up, budgets form around your category without you in the room, and customers put you in their architecture diagram without being asked. From inside any single good meeting, the two are indistinguishable. Vision-market fit is dangerous precisely because it looks like the opening scenes of product-market fit. The two are different, and scaling on top of the first one means scaling a hypothesis. The difference only shows up in what happens when you stop pushing, which is exactly the experiment a scaling company never runs. ![The map I was actually on, drawn with axes I could only name later. Vision-market fit is the top-left quadrant: resonance keeps climbing, the pull line never gets crossed, and the quadrant keeps you alive exactly long enough to matter. Rooms lean in, rounds come together, pilots start, and none of it compounds on its own.](https://scottclark.io/images/blog/make-all-new-mistakes-faster/make-all-new-mistakes-faster.png) Naming the problem is what makes it escapable. An unnamed false positive just feels like early traction running a little late. A named one is checkable: you can ask, every quarter, what moved this quarter that we didn't push, and write the answer down. In our case, the answer had been on the page for a while: the doubts that eventually proved decisive were sitting in the margins of our own strategy decks, in our own words, well before I acted on them. Strategy documents archive your disconfirming evidence whether you mean them to or not. And there's an irony worth saying plainly: I spent those years building software to catch the gap between how people assumed their AI systems behaved and how they actually behaved, but people were still figuring out these systems or just YOLOing them into production. The assumption I tested least was the one my own company was standing on: whether people had enough data, risk, and cared enough to actually test rigorously. I still believe the AI-testing market will mature the way I believed it would, but it wasn't maturing on the clock we were running. Naming the trap also names the way out. *The honest process* is what you run when the answer keeps coming back "nothing pulled": you test the company against the real market instead of against your own hopes, with real conversations, serious diligence, and real alternatives on the table. You run it not because you've already decided to sell or to stop, but because a choice made against real alternatives is a decision, and a choice made against imagined ones is a story you tell yourself. ## The accounting So here's the honest version, without spin in either direction. Distributional did real work on a real problem, with a fantastic team and investors I'm grateful for, and the plays I imported didn't convert vision into product-market fit. The job of the CEO is to make the business successful and I was burning cash without making enough progress. I can date the moment the pattern finally clicked into place for me. We were at an onsite with the team and had just had the best sales call in the company's history. They were basically pitching us the whole time, they had our problem, they were looking for our solution, they wanted to pilot as fast as possible. They shared our vision for the future. But when I got back to my room I started doing the math. If this deal went perfectly, if we hit every milestone and didn't catch any enterprise sales drag, then we could convert the pilot in 6 months and lock in our quoted rate. But I knew from SigOpt that there were dozens of things that would need to go right for that to happen, and we would need a dozen deals just like it to hit the numbers we needed, and every other deal in the pipeline was worse than this one. Even if we did great, it wasn't going to be good enough. If the best isn't good enough, you need to play a completely different game. This led to our first reduction in force and pivot, after 2 years of hard sprinting. Then a similar thing happened again after we pivoted from pre-production testing to post-production analytics. ([Read the pivot story here](https://distributional.com/blog/distributional-is-now-talaria).) So I tried to run a process to find the company a home. We ran the process instead of arguing with it, all the way to the end: term sheets on the table, an offer for the whole team. At the end of the day, that's what made what came next an intentional decision instead of a drift. We chose a hard pivot over both real alternatives: the exit on the table and the slow burn into nothing. We cleaned it up as well as we could: commitments honored, most of the team landing well together via one of the acquihire offers, and the company coming out the other side pointed at something I believe in even more. That something is [Talaria Scientific](https://talariasci.com), still under the Distributional banner, just a new brand and hard pivot. The short version of the mission: A multi-agent harness for computational science, helping computational scientists research faster, better, and cheaper with AI. ## New mistakes, on purpose What I'm doing differently this time around is not "avoiding mistakes." The goal isn't fewer mistakes; it's newer ones, made faster, against current evidence instead of old wins. Staying lean: Talaria Scientific today is a team of one plus contractors. Not because scale is bad, but because scale buys speed at the price of rigidity, and rigidity is what slowed my updating last time. Headcount converts hypotheses into commitments. I want my hypotheses cheap to kill for as long as possible, just like I did when building open source from my PhD before starting SigOpt. Being my own customer: this time the ideal user is me. I spend my days running computational research on PDE neural operators and Bayesian optimization through the product I'm building, and when it's slow or wrong I feel it the same afternoon. The feedback loop went from quarters to hours, which means my gradient information finally comes from the present instead of from 2016. Sequencing truth before scale: vision-market fit opens doors, and I'm glad for every one, but it isn't the signal to hire against. Product truth first, then pull, then scale, in that order, with the order enforced by keeping the company too small to fake it. That being said, the answer isn't to invert every play that worked before. Inverting still lets the old data drive: overfitting with a minus sign in front. Some SigOpt lessons transfer just fine. The discipline is holding the old playbook as a prior instead of gospel, weighting it by how much the world has moved, and letting present evidence make the calls. In a market moving this fast, the honest weight on eight-year-old experience is smaller than feels fair. There's a bigger question underneath all of this, about when to keep climbing the mountain you're on and when to change mountains entirely. Careers have many paths up many mountains, and that idea gets its own post. For now, the all-new mistakes are underway. I can't tell you which ones they are yet, which is the point: if I could name them, they'd be the old kind. The new map is mostly blank, but one quadrant on it is already named. In a few years Talaria Scientific will have its own list of what worked, and I'm sure I'll look back with all new lessons learned. --- Follow me at [@DrScottClark](https://twitter.com/DrScottClark) to see new posts. Source: https://scottclark.io/blog/make-all-new-mistakes-faster (Scott Clark) ================================================================ # Building this blog: Part II > How agents turned a decade-old blog backlog into a live site built for two readers: people, and the AI agents they send to read about you. Author: Scott Clark Published: 2026-07-28 Pillar: Building Things Canonical: https://scottclark.io/blog/building-this-blog-part-ii --- Twelve years ago I published a post on this domain called ["Building this blog: Part I"](https://web.archive.org/web/20141025075605/http://www.scottclark.io:80/blog/2/building-this-blog%253A-part-i). There was never a Part II. This is Part II, twelve years late, and the reason it finally exists is also the reason it might be worth your time: agents changed the economics of owning your own corner of the web, right as owning one started to matter again. When someone wants to know who you are now, they often don't search; they ask an AI, whether that is a web chat app or [Claude Code](https://claude.com/product/claude-code). A conference organizer asks a model whether you'd make a good speaker. An investor asks for a brief before a call. A hiring manager asks what you've actually shipped. The answer comes back assembled by an agent reading the public web on their behalf, and for that reader the front page isn't a search results page. It's whatever the agent can find, parse, and trust. I want those answers shaped by what I've actually written (ie by my own record rather than by whichever fragments of it leak through other people's platforms). Writing that lives only inside a walled garden is at the platform's mercy twice over: the feed decides which humans see it, and the crawler policy decides whether agents see it at all. Ask a model about someone whose whole record sits behind those walls and you get an answer stitched from press mentions and stale bios, an answer about them that they don't own. The alternative has a name in the same spirit as SEO. People are starting to call it AIO: optimizing your presence for the AI systems that read the web on people's behalf. It feels like early SEO, and the mechanics are simple: a canonical site you control, a markdown twin of every page so an agent gets clean full text instead of your layout, structured data so the answer engine doesn't have to guess, and a robots.txt that says yes. Whatever the discipline ends up being called (the SEO crowd already says GEO and AEO), the mechanical version is a few hours of work, and the agents people actually send already fetch it. Engineers have been told to own their own site for twenty years. Most of us never finish. I know because I'm most of us. ## Why this used to be a tax In 2014, while I was at Yelp, I built scottclark.io from scratch on Pyramid, because Yelp was standardizing on Pyramid and I wanted to learn it. I deployed it to DigitalOcean. I wrote exactly two posts. The second one was titled "Building this blog: Part I." What killed it was the maintenance tax. Every post required a context switch from research into web infrastructure: dependency upgrades, design, theme fights, deploy scripts. The site was supposed to be a writing surface and became a side project that ate the writing time. It was the woodshop problem: the first project in a new shop is always a workbench, and some of us sand the workbench for years and never build the furniture (and I tend to fall victim to yak shaving, bike shedding, and all the colloquial variants of this). My workbench had a comment system, but it only had two posts. So the site went dark, and the pattern held for a decade. My resume repo tells the same story in commit history: steady updates through 2016, then nine and a half years of nothing public. ## The unlock, and the new tax What changed isn't that web frameworks got easier. The frameworks have been good for years. What changed is that the scaffolding tax became an agent's problem: the dependency upgrades, the build pipeline, the infrastructure, the long tail of small decisions that used to stand between "I should write this down" and a live page. This was the same realization that led me to start [Talaria Scientific](https://talariasci.com/blog/why-im-building-talaria) to accelerate computational science research. Agents don't get you off the hook, though. They are very good at rapid prototyping, and there is still a lot of art and craft in making the prototype into the thing it was prototyping. Point an agent at "build me a personal site" and you get a generic landing page with rotating gradient blobs. What you actually owe the project is judgment (or what many people are calling taste): which tools do which jobs, what the contracts between them say, what stays open and what stays private, which way a fact propagates when it changes. The tax moved up a level, from scaffolding to composition, and composition is a much better use of the human. Three repos came out of that judgment for me. **[content_finder](https://github.com/sc932/content_finder)** is a small, still-pre-v0.1 Claude Code-driven crawler that walks the public web for a person, verifies what it finds against stable identifiers, and produces a dossier on yourself: markdown files with citations and raw captures of every primary source it can find on the internet. I ran it on myself first. The output is the closest I have to a verified record of my public existence: twelve years of articles, every podcast and talk transcript I could recover, patents, papers, and press, each with the raw artifact saved so nothing has to be re-derived. **resume** is my old LaTeX [resume repo](https://github.com/sc932/resume), modernized in a single April session after its long sleep. It now carries a markdown twin that is the source of truth for bio facts, a PDF build, and an AGENTS.md that defines the propagation order when a fact changes: dossier, then resume, then site, never the other direction. **scottclarkio** is the published surface [as a repo](https://github.com/sc932/scottclarkio): an Astro static site behind S3 and CloudFront, provisioned by Terraform, deployed by a GitHub Action on every push. It reads facts from the other two repos. And it ships the full agent-facing surface: a JSON-LD entity graph on every page, a markdown twin of every page and post, an [llms.txt](https://llmstxt.org) and a concatenated llms-full.txt, a full-content RSS feed, and a robots.txt that explicitly welcomes AI crawlers by name, with a Content-Signal line that says yes to search, yes to AI input, yes to training. Building in the open means being fine with getting baked into the models, and getting baked in is the goal, not a concession: whether a model learns about me at training time or looks me up at answer time, the record it finds should be the one I wrote. ![A site written for two readers. The dossier repo holds verified public facts, the resume repo holds bio truth, and the site composes both into two surfaces: HTML for the person, and a machine-readable stack (markdown twins, an entity graph, llms.txt, a full-content feed) for the agent they sent. The connective tissue between repos is markdown contracts. Facts flow one direction.](https://scottclark.io/images/blog/building-this-blog-part-ii/fig1-two-readers.png) The composition lesson is in the boundaries. Each repo does one job, none of them tries to be a CMS, and the connective tissue is markdown: CLAUDE.md and AGENTS.md files that tell the next agent session what's true, what to read first, and what not to touch. When a fact about me changes, I know exactly which repo to edit, and the rest follow. There is no client-side JavaScript on the site; the only script tags carry JSON-LD data. There's no analytics tracker either, just the CDN's own logs: watching a live dashboard is fun, and fun isn't signal; I don't need to peek at an experiment on a minute-by-minute basis. ## What I didn't build The list of what I didn't build matters more than it used to, because the old constraint no longer enforces it. I built content_finder because the artifact didn't exist: there was no off-the-shelf way to produce a verified, citable corpus of my own public footprint, so the only path to the dossier was the tool. Halfway through drafting a post, I also started designing a markdown collaborative-editing tool in my head, because pair-editing prose with an agent in a terminal has real friction. Then I caught myself: editors exist, and the artifact I was trying to ship was the post. When the marginal cost of building a tool approaches zero, "can I build it" stops being the question, and "should I build it for this artifact, in this moment" is the question that remains. That's an explore/exploit tradeoff at the tool layer, the same one I spent years on in Bayesian optimization at SigOpt. Exploit, ie use what exists, is still the right default for almost everything, but the reason changed: it used to be that building (exploration) was too expensive, and now it's that your judgment is the scarce input and the backlog of artifacts is long. I built content_finder because I needed the dossier. I didn't build the editor because I just needed the post and Google Docs does just fine. ## It compounded The site turned out to be the trailhead, not the destination. This July, the same composition pattern shipped two more sites in one night: a company site with its own blog for [Talaria Scientific](https://talariasci.com), the company I'm building now, and a rebuilt [distributional.com](https://distributional.com) that restored years of the old company's technical archive at the original URLs. Every post on all three ships with the same agent-facing surface this site has. The same night, this site got the blog section you're reading, which is how a post titled Part II finally acquired a place to exist. The build discipline scaled with it. Before these posts go live, the diffs are reviewed by frontier models from five different labs (Anthropic, OpenAI, xAI, Moonshot, and Thinking Machines), and the union of their findings, most found by only one of them, becomes the fix list. How that review loop works is a post of its own; the short version is that agents wrote the estate, different agents attacked it, and I adjudicated. ## The speed The receipts are what still surprise me. The resume repo is the cleanest one: regular commits through October 2016, then silence, then a single commit in April 2026 titled "Modernize resume; align CV/md; add AGENTS.md and fork guide" that did more than the previous five years combined. When the company [pivoted to Talaria Scientific](https://distributional.com/blog/distributional-is-now-talaria), the resume, the CV, the PDF, and the site all updated the same afternoon, because a fact now has one home and agents carry it everywhere else. This site went from first commit to live behind CloudFront in about ten weeks of part-time sessions, including the AWS infrastructure, the DNS migration, and the whole agent-facing surface. The AIO endpoints are the part I would never have gotten to alone; JSON-LD schema authoring is exactly the kind of yak-shaving that kills a research week, and it's exactly what an agent eats without complaint. On a normal weekday I have a handful of agent sessions open, and the one working on this site is rarely the one I'm watching. At the end of the day, the work I did is the work that was supposed to be there all along: every design decision, every fact, every editorial call, every judgment about what not to build. What the agents ate was the tax that used to stop the project before that work could start. In 2014 I wrote "Building this blog: Part I" and then went quiet for twelve years. This is Part II. It didn't feel 10x faster to make; it felt 100x faster, and somewhere in that gap is escape velocity: the difference between a backlog item that survives another decade and the post you're reading right now. Part III should take less than twelve years. If you've got a workbench of your own gathering dust, it might not need another decade for you to own your slice of the web too. --- Follow me at [@DrScottClark](https://twitter.com/DrScottClark) to see new posts. Source: https://scottclark.io/blog/building-this-blog-part-ii (Scott Clark)