For our third Benchtalks, the series dedicated to the researchers building the measurement toolkits that frontier labs hill-climb on, Snorkel AI co-founder Vincent Sunn Chen sat down with Parth Asawa, a UC Berkeley PhD student advised by Matei Zaharia and Joey Gonzalez, and the creator of Continual Learning Bench, the first standardized benchmark for measuring whether AI systems actually learn from experience over time.

This interview covers:

• Why your coding agent re-reads your entire codebase from scratch every single session, and why a human developer stops doing that after task one
• How the field got so good at scaling that it quietly set aside the original question: can a model actually learn on its own after training ends?
• The poker task: if your opponent calls every hand, a human exploits it in minutes, and models run in separate sessions don't figure it out
• The gain metric: why cumulative reward alone can't tell you if a system is learning or just smarter to begin with, and how stateful minus stateless performance isolates actual learning
• What Anthropic's Fable release revealed, including how they used Continual Learning Bench to show Fable outperformed systems using Opus or Sonnet as a backbone
• Why in-context learning still leads the leaderboard, and why Parth's long-run bet is on parametric systems
• The false dichotomy in AI safety: unsafe open models vs. consolidated power, and what Parth and Joey Gonzalez propose as the third path

Full interview/transcript: snorkel.ai/blog/benchtalks-parth-asawa-continual-learning-bench/

For our second Benchtalks, the series dedicated to the researchers building the measurement toolkits that frontier labs hill-climb on, Snorkel AI co-founder Vincent Sunn Chen sat down with John Yang, a Stanford PhD student and creator of the SWE-bench franchise, SWE-smith, CodeClash, and most recently ProgramBench.

This interview covers: 

• Why every frontier model scored 0% at launch — until GPT-5.5 cracked the first task (cmatrix)
• Why ProgramBench grades the runnable artifact, not the implementation — and lets models build in any language
• The post-training tell hiding in plain sight: how much models (especially GPT) love Python, even when it looks like a handicap
• The reward-hacking problem — models with internet access cheated up to 36% of the time, and why nine LLM judges still couldn't agree on what counts as cheating
• The lineage from SWE-bench to SWE-smith to CodeClash, and what ProgramBench needs from expert contributors to grow

Full interview/transcript: https://snorkel.ai/blog/benchtalks-john-yang-programbench/

In this inaugural episode of Benchtalks, Snorkel AI co-founder Vincent Chen sits down with Alex Shaw, MTS at Laude Institute and co-creator of Terminal-Bench, to unpack what the rapid hill-climbing on TB2 reveals about the state of AI agent evaluation — and where the field needs to go.

This interview covers: 

• Why TB2 went from 20–30% during development to 75–80% at the frontier today
• The bet on the terminal as the right abstraction for general computer use
• How Harbor became a benchmark factory — and why that matters for RL post-training
• The "benchmaxxing" problem and how the community is keeping TB2 honest
• What Terminal-Bench 3 needs from expert contributors to shape model development for the next year

Full interview/transcript: https://snorkel.ai/blog/benchtalks-alex-shaw-terminal-bench-harbor-building-the-benchmark-factory/