HeyJaredMachine Learning Street Talk
MLST is the leading highly technical AI podcast. Subscribe now! Welcome! We bring you the latest in advanced AI Research, Science and Philosophy. Our approach is unrivalled in terms of scope and rigour.
- Indexed videos, last 90 days
- 24
- Latest publication
- Sep 29, 2026
- Audience
- ~220K subscribers
- Earliest in this view
- Jul 13, 2026
Latest videos
AI Can Write the Proof. Who Checks It? — Leonardo de Moura (opens the original)
Read excerpt
Leonardo de Moura created Lean and co-created Z3. --- This episode is sponsored by Parallel. Parallel, where agents find answers: web search, extraction and deep research APIs built for AI agents. Start free with the Parallel MCP server and $5 of credits every month: https://parallel.ai/mlst?utm_source=creator&utm_medium=podcast&utm_content=MLST --- Tim Scarfe talks with Leo about how Lean escaped its original audience, why dependent types and Mathlib made it useful to working mathematicians, an
Why Deep Learning Failed on Tables for a Decade - Frank Hutter (opens the original)
Read excerpt
Frank Hutter, co-founder of Prior Labs, on why deep learning struggled with tabular data for a decade: tables are messy and heterogeneous, hyped models like TabNet did not generalise to new datasets, and there was no ImageNet of tables. The breakthrough came from learning to transfer at the level of patterns across many different tables; TabPFN-3.5 now tops the TabArena benchmark. Full interview on MLST: https://www.youtube.com/watch?v=72Im-Mm5JKs References: TabNet (Arik & Pfister): https://arx
Reward hacking… or just fixing a bug? #podcast (opens the original)
Read excerpt
Weco AI's research agent wrote a gigantic monkey patch to their eval script. Reward hacking? No: it was fixing a bug. Zhengyao Jiang (co-founder and CEO, Weco AI) on why it's getting harder and harder for humans to tell the difference as agents become more sophisticated. From our episode with Zhengyao Jiang on AIDE and AI systems that improve themselves. #AI #AIAgents #RewardHacking #MachineLearning #Shorts
Can Rewriting an AI Agent Bend the Intelligence Curve? - Zhengyao Jiang (opens the original)
Read excerpt
Weco let an AI coding agent rewrite the harness around another agent for eight days: its code, prompts and tools, while the underlying language model stayed fixed. Tim Scarfe asks Weco co-founder Zhengyao Jiang what the reported gains over two years of human engineering actually demonstrate. The discussion examines AIDE 85's generated code, held-out evaluation and the difficulty of separating useful discoveries from reward hacking. Jiang explains Weco's four levels of recursive self-improvement
A perfect model can still get causality backwards (opens the original)
Read excerpt
Frank Hutter, co-founder and CEO of Prior Labs, on why a model that predicts perfectly can still point you to the wrong decision. If patients who get a medicine tend to have a disease, stopping the medicine will not cure it. Observing is not the same as intervening, and that difference is what causal machine learning is for. From our conversation with Frank on TabPFN and foundation models for tabular data. #Shorts
Publishing over time
Last 90 days. Choose a month to open its work.
Recurring subjects
Named in the text we hold. One piece can cover several.
Audience
~220K subscribers
Measured Sep 19, 2026
Source's subscribers, not the number who saw an individual piece.
About this data
Counts cover the work we have indexed. Tone needs enough text and a confident classification. Excerpts and episode notes are not full articles or transcripts.
Identity or attribution wrong? Suggest a correction.