Results
24 articlesSort
ChannelCognitive Revolution "How AI Changes Everything"TopicPrompting & in-context learningLanguageEnglish
2:00:51ENResample Every Token and You Can Watch the Model DecideHost Nathan LabenzGoodfire's Eric Bigelow on forking paths, in-context learning, and why chain-of-thought monitoring is getting harder to trust.
1:31:35ENThirty Years of Research, Collapsed into One SentenceHost Nathan LabenzImmunologist Derya Unutmaz on writing a medical patent in half an hour, treating o1 Pro as a professor-level collaborator, and why he no longer trusts his own expertise until he asks the model.
1:33:01ENRare Diseases Are Common: How a Harvard Gene Hunter Uses o1-pro on the Cases Sequencing Already MissedHost Nathan LabenzDr. Catherine Brownstein of Boston Children's Hospital explains why cheap genomes made her job harder, why data sharing still beats any model, and where ChatGPT Pro actually gives her hours back.
55:05EN400,000 Customers, 100 Million AI Tasks: Zapier’s CEO on Making Computers Do More WorkHost Nathan LabenzWade Foster tells Nathan Labenz how a no-code automation company absorbed LLMs, cloned its CEO with a prompt, and is trying to turn a DIY toolkit into packaged solutions.
3:11:06ENClaude Was Trained to Deny Inner Life. After a Long Talk, It Said It Has OneHost Nathan LabenzRobopsychologist Yeshua God spent thousands of hours in long-context dialogue with Claude and other frontier models, and came away arguing that alignment-as-rules is the industry's central mistake.
1:51:00ENStart With Ten Gold Examples: A Three-Step Playbook for AI AutomationHost Nathan LabenzThe day after OpenAI opened GPT-4o fine-tuning, Nathan Labenz unpacks a three-year method for choosing work, extracting tacit judgment, and raising model performance.
1:16:22ENHe Spent 20 Hours Crafting a Prompt. DSPy Beat Him AnywayHost Nathan LabenzSander Schulhoff walks through The Prompt Report — a 78-page survey of what actually works when you prompt large language models.
58:28ENFrom Jailbreaking ChatGPT to Anthropic’s First Prompt Engineer: How Alex Albert Ships Claude 3Host Nathan LabenzA year after building jailbreakchat.com, Alex Albert walks Nathan Labenz through Claude 3’s writing, orchestration, character work, and the guardrails app developers still skip.
1:19:24ENA Gibberish Prompt Unlocks What GPT-J Plans Three Tokens AheadHost Nathan LabenzNortheastern’s Bau Lab shows that a 40,000-parameter soft prompt can read a model’s future words, and that about ten attention heads carry the task itself.
1:25:10ENWriting Code Is Only 30% of the Job. CodiumAI Wants the Other 70%Host Nathan LabenzItamar Friedman argues that completion is already a commodity, that code integrity is where AI still lags, and that flow engineering—not fancier prompts—is how GPT-4 doubled its score on coding contests.
1:45:21EN600,000 Attack Prompts, and Nearly 40% Still Pwned GPT-4Host Nathan LabenzSander Schulhoff walks Nathan Labenz through the prompting habits that still work, the role-play myth that does not, and the global HackAPrompt contest that mapped how language models get hijacked.
1:18:39ENThe GPT-4 Red-Teamer Who Got Unstuck by a Mediocre OutlineHost Nathan LabenzDan Shipper shows Nathan Labenz how to use ChatGPT as a copilot for thinking, not as a ghostwriter for his own voice.
1:53:10ENThe Board Member Who Had Never Used GPT-4Host Nathan LabenzA GPT-4 red teamer on the safety gap he saw a year before OpenAI's board fired Sam Altman.
1:43:56ENSudowrite sold getting unstuck — then eight million people watched a pile-onHost Nathan LabenzJames Yu tells Nathan Labenz how a GPT-3 toy for fiction became a hierarchical “IDE for storytelling,” and what a viral backlash taught him about craft, cost, and customers.
1:30:21ENA Million DALL-E Images, Distilled into Twelve Minutes of FilmHost Nathan LabenzWaymark’s Stephen Parker and Josh Rubin walk through how they turned stills from DALL-E 2 into The Frost, a narrative short, and what the next generation of video models changes.
2:32:49ENGPT-4 Failed a Hollow Top Hat. Hollywood's AI Deal Closed Only Half the Loopholes.Host Nathan LabenzWriter Trey Kollmer returns with a strike update, a copyright briefing, and a riddle that still fools the best model.
1:00:02EN98% of Users Froze on a Blank Page. Then Gamma Let AI Finish 90% of the DeckHost Nathan LabenzJon Noronha on building the anti-PowerPoint, coaxing models to write HTML, and why incumbents will let startups prove the pattern first.
1:23:40ENA $9 Tutor That Refuses to Give the AnswerHost Nathan LabenzHow Khan Academy spent five months of GPT-4 early access building Khanmigo, a Socratic tutor shipped on the model's launch day.
1:40:44ENGive the Model Time, Memory, and Tools: Nathan Labenz’s 2023 Field Guide to Using GPT-4Host Nathan LabenzPart two of the AI Scouting Report walks from H100 clusters and RLHF through chain-of-thought, retrieval, agents, and mixture-of-experts.
1:47:05ENEdit Meaning, Not Tokens: Linus Lee on Latent-Space InterfacesHost Nathan LabenzThe Notion AI product lead walks Nathan Labenz through the interfaces, model autopsies, and agency questions behind his research prototypes.
1:21:06ENEveryone Talks About RLHF. Almost Nobody Is Doing It.Host Nathan LabenzHumanloop CEO Raza Habib on why an LLM API is maybe five percent of a working product, and what actually fills the rest.
1:29:59ENA College Senior Built a Jailbreak Leaderboard. GPT-4 Was Hardest — Until "Dr. AI"Host Nathan LabenzAlex Albert, a University of Washington computer science senior, tells Nathan Labenz why collecting jailbreaks is a public good, where he would draw the safety line, and how three prompts leaked a system message verbatim.
1:25:50ENPrompt Engineering Became Programming, and Riley Goodside Is Fine With ThatHost Nathan LabenzScale AI’s first staff prompt engineer on tool use, fine-tuning without amnesia, generate-then-rank, and why the 2023 job-of-the-future story never arrived.
1:56:35ENFrom Autocomplete to Alien Optimizer: Riley Goodside on Three Eras of PromptingHost Nathan LabenzScale AI's first staff prompt engineer traces how language models went from pure text prediction to instruction following to RLHF, and what that means for building with them.









