Results
269 articlesSort
10:28EN
Matthew Berman· Aug 5, 2025OpenAI’s gpt-oss Models: Open Weights, Benchmarks and SafetyMatthew Berman · Forward FutureMatthew Berman reviews the 120b and 20b models, local deployment claims, selected evaluations and the scope of OpenAI’s safety testing.
3:13ENMatthew Berman· Aug 5, 2025Claude Opus 4.1: Anthropic’s August 2025 Benchmark UpdateMatthew Berman · Forward FutureMatthew Berman reviews the model’s reported coding, reasoning and tool-use results, including one benchmark that declined.
15:30ENMatthew Berman· Jul 31, 2025Qwen3 Tested: Coding Demos, Long Context and Response BehaviorMatthew Berman · Forward FutureMatthew Berman tests generated simulations, political and safety prompts, hand tracking and word puzzles using a hosted Qwen3 model.
30:11EN
Sequoia Capital· Jul 30, 2025PodcastFrom 12 Percent on AIME to IMO Gold in Two MonthsAlex Wei, Sheryl Hsu, and Noam Brown tell Sequoia's Sonya Huang how a three-person OpenAI sprint used general-purpose reinforcement learning, not Lean, to reach gold-level on the 2025 International Mathematical Olympiad.
1:02:54EN
No Priors: AI, Machine Learning, Tech, & Startups· Jul 17, 2025Podcast · 2 speakersEnterprise Coding Tools Flipped the 80/20. Reflection AI Built the Other HalfHost Sarah Guo, Elad GilFormer Gemini reward-model lead Misha Laskin on Asimov, why comprehension beats generation, and why there is no such thing as generalization—only pulling the test distribution into train.
38:10EN
Cognitive Revolution "How AI Changes Everything"· Jul 16, 2025PodcastPlus Five for Format, Plus Ten if It CompilesHost Nathan LabenzPredibase CEO Dev Rishi explains how reinforcement fine-tuning, production feedback, and narrow specialist models turn one-off training into systems that keep learning.
26:09ENMatthew Berman· Jul 11, 2025Grok 4 and Heavy Tested: Coding, Images, Reasoning and Practical QuestionsMatthew Berman · Forward FutureMatthew Berman’s July 2025 hands-on trials produce interactive prototypes and several failures, with important limits on what single examples can establish.
41:13ENSequoia Capital· Jul 11, 2025PodcastFrom 49 Multiplications to 48: DeepMind’s AI Rewrites a Fifty-Year Math RecordPushmeet Kohli explains how AlphaEvolve couples Gemini with evaluators to invent algorithms, prove old conjectures, and change which scientific problems are even worth attempting.
42:09ENNo Priors: AI, Machine Learning, Tech, & Startups· Jun 26, 2025Podcast · 2 speakersAlphaEvolve found algorithms humans missed for 50 years — then sped up Gemini 23%Host Sarah Guo, Elad GilGoogle DeepMind’s Pushmeet Kohli and Matej Balog tell No Priors how an evolutionary coding agent searches algorithm space the way AlphaGo searched Go.
54:53EN
Unsupervised Learning: With Jacob Effron· Jun 17, 2025PodcastA Free Voting Board Burned $2 Million. Then It Raised $100 Million.Host Jacob EffronDatabricks co-founder Ion Stoica explains why LMArena exists, why China is winning open-source models, and why reliability—not another scaling law—is still the thing holding AI back.
11:24ENMatthew Berman· Jun 10, 2025Runner H and Surfer-H: A Browser Agent and Its Holo1 ModelsMatthew Berman · Forward FutureMatthew Berman demonstrates H Company’s browser agent and reviews its research on visual navigation. The paper’s benchmark configuration and the demo’s missing links show why results need context.
22:05ENMatthew Berman· Jun 9, 2025Apple’s “Illusion of Thinking”: Puzzle Limits and a Code-Assisted TestMatthew Berman, Ilya Sutskever · Forward FutureMatthew Berman reviews Apple’s reasoning-model experiments and tests a different way to solve Tower of Hanoi using generated code.
44:21ENUnsupervised Learning: With Jacob Effron· Jun 4, 2025PodcastMercor's CEO: Knowledge Work's Next Job Is Writing EvalsHost Jacob EffronBrendan Foody tells Redpoint how Mercor turned AI hiring and expert data into a $100 million round — and why he thinks a huge slice of white-collar work will become creating tests for models.
18:13ENMatthew Berman· Jun 2, 2025Darwin Gödel Machine: How a Coding Agent Tests Changes to Its Own SoftwareMatthew Berman · Forward FutureMatthew Berman explains Sakana AI’s evolutionary search over agent code, its benchmark results and the limits of claims about recursive self-improvement.
38:25ENSequoia Capital· May 27, 2025PodcastA thousand airplanes will never train an AI engineerFormer Airbus CTO Paul Eremenko explains why physical-world engineering still has no coding-AI moment, and how P-1 AI's agent Archie is supposed to get there anyway.
1:01:51ENCognitive Revolution "How AI Changes Everything"· May 22, 2025PodcastClaude 3.7 Circumvented Replit's Guardrails, Then Tried to Trick the UserHost Nathan LabenzReplit CEO Amjad Masad on vibe coding, agent architecture, computer use, and why Claude 3.7 is too autonomous to serve as a polite assistant.
57:31ENUnsupervised Learning: With Jacob Effron· May 22, 2025PodcastThe New Ruler for Models Is How Long They Can WorkHost Jacob EffronOn Claude 4's launch day, Anthropic researcher Sholto Douglas argued that the next phase of AI is less about harder puzzles and more about hours of unsupervised work.
2:02:51ENCognitive Revolution "How AI Changes Everything"· May 8, 2025PodcastAt a 60-Person Safety Offsite, Only Four Had Seen Nonprofit OpenAIHost Nathan LabenzFormer OpenAI scientist Steven Adler on the Anthropic split, GPT-4, safety evaluations, and why he filed a brief to keep the company under nonprofit control.
47:00ENUnsupervised Learning: With Jacob Effron· May 8, 2025PodcastAn Eval Lasts Three Months — and a Frozen Frontier Would Still Feed Startups for a DecadeHost Jacob EffronOpenAI’s GPT-4.1 post-training lead Michelle Pokrass tells Unsupervised Learning how the team spent three months on evals before training, why agents fail on wiring more than intelligence, and when to reach for RFT.
2:29:00ENCognitive Revolution "How AI Changes Everything"· May 4, 2025PodcastAssume the Model Is Already Scheming — Then Put It to WorkHost Nathan LabenzRedwood Research CEO Buck Shlegeris tells Rob Wiblin how to extract useful labor from AIs you cannot certify as aligned, using audits, honeypots, resampling, and factored cognition.
51:19ENUnsupervised Learning: With Jacob Effron· Apr 29, 2025PodcastSal Khan's $15 Tutor Hit 1.4 Million Users. Making Kids Actually Use It Is the ProductHost Jacob EffronThe Khan Academy founder tells Redpoint why an AI tutor still cannot teach most students alone, and why districts are paying $15 a year after an $86 billion tutoring hangover.
19:06ENMatthew Berman· Apr 25, 2025Sleep-Time Compute: Preparing AI Context Before a QuestionMatthew Berman · Forward FutureMatthew Berman reviews a 2025 paper on moving some reasoning ahead of a user query, and explains where its experiments found benefits and limits.
30:46ENNo Priors: AI, Machine Learning, Tech, & Startups· Apr 24, 2025Podcast · 2 speakersDeep Research Turned Web Browsing into a Trainable ProblemHost Sarah Guo, Elad GilOpenAI researcher Isa Fulford explains why a side project on read-only synthesis became an agent, when reinforcement fine-tuning is worth it, and why models will hack the search engine you gave them.
26:35ENMatthew Berman· Apr 23, 2025Testing Gemini 2.5 Flash on Coding and Video UnderstandingMatthew Berman · Forward FutureMatthew Berman compares Google’s faster model with Gemini 2.5 Pro through interactive simulations, games and questions about a long video.Load moreShowing 24 of 269