1st10 Podcast

When a chatbot says something harmful, who do you hold responsible? Who gets to set their boundaries, and who answers when those choices go wrong?On this episode of the 1st10 Podcast, Boris Epstein speaks with Suzanne Nossel, former CEO of PEN America and a member of Meta's Oversight Board, about speech, safety, and the power behind platform rules. Suzanne Nossel sits on Meta's Oversight Board, where she rules on what can stay online. Boris and Suzanne's conversation examines a key question around AI that is dominating discourse today: Now that machines do the talking, who should set the limits for AI?They discuss, among other things:How the Oversight Board weighs expression against other rights, What power the Oversight Board lacks that makes the board weaker than it looksWhy any labels meant to clarify AI-made content could backfire heavily and make matters worseWhy the legal shield that protected social platforms may not apply to AI companiesWhy a founder's AI choices can reach into customer trust, liability, and public acceptance.What Dario Amodei must do to ensure Anthropic's new embedded evaluators have real teeth. As CEO of PEN America, Suzanne Nossel spent years ruling on what people are allowed to post. Now she faces a harder version of the same question: when the speaker is software and has no human author who can lay claim to its words, who gets held responsible for what it... says?Chapters:00:00 Preview Of The Episode00:41 Introducing Suzanne Nossel01:50 A Career That Resists a Label05:39 The Tweet That Started a Fight07:39 When The Agents Surprised Everyone12:37 The Trust Problem Around Ai15:23 The "AI Labeling" Problem20:36 Social Media's Messy Report Card25:19 A Second Opinion for Anthropic27:20 One Lesson for Every AI Startup31:26 What If the Labs Get It Wrong?34:42 What If the Labs Get It RIGHT?!39:59 What Her Law Students See44:44 Is Anyone Getting It Right?46:09 A Bold Bet on 202747:05 How to Reach Suzanne NosselQuotes:"I don't have a lot of faith that there's going to be a robust regulatory regime put into place anytime soon." - Suzanne Nossel (12:10)"When you unleash these powerful technologies, the results really are unpredictable, and there will be things that you never anticipated." - Suzanne Nossel (27:30)"No human being decided that that was going to be the bot's response. What the human beings did was configure a bot that, I think unbeknownst to them, would and did yield that response." - Suzanne Nossel (30:39)"What was completely missing were any user rights or any hard constraints on the model. [...] To me, it read like a parent's letter to their baby child." - Suzanne Nossel (45:41)Follow:Spotify: https://podcasters.spotify.com/pod/show/1st10podcast Amazon: https://music.amazon.com/podcasts/7e8ec9af-f38c-4cd9-8c68-1c1dd4516b 27/1st10-podcastApple: https://podcasts.apple.com/us/podcast/1st10-podcast/id1760411207 Podcast: https://www.1st10.com/podcast RSS: https://anchor.fm/s/f951319c/podcast/rss YouTube: https://www.youtube.com/@1st10podcast Twitter www.x.com/1st10engineersLinkedIn: www.linkedin.com/company/1st10/Links:Suzanne Nossel on LinkedIn: https://www.linkedin.com/in/suzanne-nossel-0118011b/Suzanne’s Profiles:Wikipedia: https://en.wikipedia.org/wiki/Suzanne_NosselOversight Board: https://www.oversightboard.com/people/nossel-suzanne/PEN America: https://pen.org/profile/suzanne-nossel/Global Affairs: https://globalaffairs.org/research/experts/suzanne-nosselForeign Policy: https://foreignpolicy.com/author/suzanne-nossel/Music by Roman Senyk from PixabayProducer: Shrikant Joshi

Philip's three-word warning drove this channel's most commented video. People are still arguing about it. Now he is back. On this episode of the 1st10 Podcast, host Boris Epstein sits down with Philip Su and asks him about his prediction about winter coming for developers that got 900k views (and counting!) over the past year. And ONE day before Philip starts as Engineering Manager at Anthropic Labs, we ask him to score his own call. This time around, Philip revisits it with more nuance: junior hiring is under real pressure, but AI is also making ambitious engineers more capable than ever. He explains why the youngest engineers are already feeling the cold while everyone else is not, why AI is splitting developer pay into two very different tiers, and why cheap code demands more discipline, not less. Tune in to hear Boris and Philip talk about:The score Philip gives his own prediction and the one group he says is already in the AI Winter.The skills that entry-level engineers now need. (HINT: They used to belong to two other jobs!)The change in a manager's job when every contributor has a small 'team' working beside them.The safeguard that teams lose when code costs almost nothing to write and the discipline that has to replace it.A new plane-boarding habit that explains why AI won't give engineers their evenings back.A small law in New Jersey is Philip's model for how AI politics *might* go wrong.Philip does not offer a safe prediction this time around. Instead, he offers a clearer question: when the environment churns, will you protect the old map or learn to move before everyone else does?Chapters:00:00 Preview of the Episode01:42 So, Did Winter Come? Who for?06:54 Philip Scores His Own Prediction 08:28 When CS Paid Less Than Median Wage12:11 Code Is No Longer The Bottleneck?16:50 A different way to think about AI risk20:06 Why Philip Is Still Optimistic23:01 Philip'S Next Chapter: Anthropic Labs27:48 Managing People or Agents? (Or Both?)33:05 The New Job Description Of A Great Developer38:21 Non-AI Career Advice From Philip45:22 Ai As A Political Question48:58 Shovels, Spoons, and New Jersey?!53:05 Philip Su for... Political Office?!53:39 How to Contact Philip SuQuotes:"People in software industries who are entering industry now have a fifth fewer jobs than they used to." - Philip Su (03:09)"I do think we will once again enter an era where you should only go into computer science because you really love the work." - Philip Su (08:28)"With AI, what you realize is that there is never downtime for you." - Philip Su (14:23)"A team without discipline simply produces more things that they toss more out of because they don't feel bad tossing it out." - Philip Su (31:52)"The type of person who can't have their cheese moved is going to suffer a lot." - Philip Su (42:21)Connect:Spotify: https://podcasters.spotify.com/pod/show/1st10podcast Amazon: https://music.amazon.com/podcasts/7e8ec9af-f38c-4cd9-8c68-1c1dd4516b 27/1st10-podcastApple: https://podcasts.apple.com/us/podcast/1st10-podcast/id1760411207 Podcast: https://www.1st10.com/podcast RSS: https://anchor.fm/s/f951319c/podcast/rss YouTube: https://www.youtube.com/@1st10podcast Website: www.1st10.comTwitter www.x.com/1st10engineersLinkedIn: www.linkedin.com/company/1st10/Links:Philip Su on LinkedIn - https://www.linkedin.com/in/suphilip/Philip's Substack: https://molochinations.substack.comProducer: ⁠Shrikant Joshi⁠ (https://42quirks.com)Fact-checks:U.S. Bureau of Labor Statistics, AI and employment projections - https://www.bls.gov/opub/ted/2026/artificial-intelligence-information-technology-and-employment-2024-34.htmNo Widespread Displacement, but the AI Employment Gap for Young Workers Has Widened to 19% - Stanford Digital Economy Lab - https://digitaleconomy.stanford.edu/news/canariesaug26/Employment Situation News Release - 2026 M09 Results - https://www.bls.gov/news.release/empsit.htm#ces_table1.f.p

The hardest part of leaving Big Tech is the certainty and stability of Big Tech. Kobi Hudson spent his entire adult life at Waymo, and he still left to join a startup.On this episode of the 1st10 Podcast, Boris Epstein sits down with Kobi Hudson, Founding Engineer at Coval.ai and asks him the question that every experienced software engineer has on his mind: When does leaving Big Tech makes sense for a great engineer? And when it does not? Boris and Kobi discuss what changes when an engineer moves from a large, highly resourced organization to a clean-slate startup, including:The one thing about his Waymo job that kept changing until he started looking elsewhere.How a founder, a lunch, and about a month of conversations made a leap from a stable job feel survivable.The tool switch that broke 17 years of coding habits and how it symbolizes life outside Big Tech.The liberating feeling afforded by a clean slate that frees you from the burden of carrying years of tech-debt.Theglaring bugs that startups are willing to leave alone because ONE question that decides which bugs get fixed.The amazing developments currently happening at the bleeding-edge of Voice AI - systems that can listen, think, speak, interrupt, and use tools in more natural ways.ANDWhy a tomato from your own garden is a perfect metaphor for local AI models.Kobi swapped his Big tech job of proving autonomous systems were safe for a startup job built around proving AI conversations trustworthy. His career move, when you think about it, is a test of the same engineering instinct: you must know what is true before you let the system loose.Chapters00:00 Preview of the episode00:30 What is Kobi Hudson All About?04:03 Designing mobility for the world08:02 Culture Shock, Courtesy California11:37 What Made Leaving Less Scary15:16 What Kobi Inherited As a Founding Engineer19:30 Knowing The Founder Certainly Helps 22:13 The Culture Shock Kobi Didn't Expect25:20 A Test for Anyone on the Fence27:33 Bleeding Edge of Voice AI32:49 Coval Is (And Will Continue) Hiring34:33 Kobi's Tomato Theory for AI36:39 Kobi's Predictions For The FutureQuotes:"I wanted to be part of growing a team, and I wanted to be part of shaping the culture." - Kobi Hudson (13:33)"When I switched, I started using Cursor. And so basically it felt like going from zero to 100." - Kobi Hudson (22:47)"I think the biggest takeaway is you've got to be comfortable with being uncomfortable." - Kobi Hudson (25:33)"When you grow a tomato in your garden and the tomato is not necessarily better than the tomato at the store, but it's your tomato. So it tastes a little bit better." - Kobi Hudson (36:00)Follow:Spotify: https://podcasters.spotify.com/pod/show/1st10podcast Amazon: https://music.amazon.com/podcasts/7e8ec9af-f38c-4cd9-8c68-1c1dd4516b 27/1st10-podcastApple: https://podcasts.apple.com/us/podcast/1st10-podcast/id1760411207 Podcast: https://www.1st10.com/podcast RSS: https://anchor.fm/s/f951319c/podcast/rss YouTube: https://www.youtube.com/@1st10podcast Twitter www.x.com/1st10engineersLinkedIn: www.linkedin.com/company/1st10/Links:Kobi Hudson LinkedIn:https://www.linkedin.com/in/kobi-hudson-998055abGitHub: https://github.com/kobicovaldevCoval: Website: https://www.coval.ai/about-usLinkedIn: https://www.linkedin.com/company/covaldev[Other Links]Announcement of the Hiscock Memorial Award for “Growing Algae in Space: https://www.nasa.gov/wp-content/uploads/2015/03/2014_mt_space_grant.pdfMusic by Roman Senyk from PixabayProducer: Shrikant Joshi (https://42quirks.com)

What happens when an AI agent gets permission to rewrite the code that runs it?On this episode of the 1st10 Podcast, host Boris Epstein sits down with Alex Krentsel, a senior researcher at Google Research and a systems PhD student at UC Berkeley to dig into why efficiency, and not accuracy, may be the real unlock for self-improving agents. Alex is building Exo, an open-source agent harness that watches itself work and changes its own policy to get better and cheaper at a task. Together, Boris and Alex talk about:The multi-billion dollar reason today's LLM providers secretly have zero incentive to make their agent loops efficient.The statistical reality behind why agents still fall apart on long-horizon corporate workflows.The unexpected childhood obsession that provided Alex with a foundational blueprint for understanding distributed agentic systems.Alex's unpopular opinion about which part of the classic ML curriculum is now "far over-weighted."The truth behind why a widely popular tech-career shortcut from the 2010s has officially ceased to exist today.Alex spent his early days engineering high-traffic UI at YouTube. These days, he works at the intersection of systems research, agentic AI, and the changing role of software engineers. The conversation ultimately comes back to a simple question, that owes its existence to Alex's childhood fascination with... ants. "How do simple components, following the right rules, become systems that are greater than the sum of their parts?"Tune in for an unfiltered look at what it truly takes to build software on the frontier of AI!Chapters00:00 Preview of the episode01:54 Meet Alex Krentsel03:16 The Architectural Shift Beyond LLM Scaling06:48 Inside Exo: AI Agents That Improve Themselves10:27 Why Exo Is Open Source13:08 How Smart Are AI Agents, Really?16:03 Explosive Adoption of AI Coding Agents18:06 What Will Software Engineers Do Now?21:34 Machine Learning v/s Systems Thinking23:36 What Should Startups Look for in Engineers?27:31 A Peek Inside Google Research31:34 Do All Roads Lead to AI?34:23 Anthills, Agent Swarms, and "Civilizations"37:17 Alex Is Optimistic About AI40:33 The Mood Among CS PhDs (at Berkeley)43:45 Contact Details and First Contact Message44:19 A Prediction for the FutureQuotes"Claude Code and Codex have zero incentive to be efficient for you, because you're paying for the token use. The more expensive it is to do the task, the more they get paid." - Alex Krentsel (08:27)"Our jobs are no longer to do the rote work of programming a UI. A lot of what we're doing now is reasoning about specification." - Alex Krentsel (18:48)"I really think that the systems skillset ... is going to be the valuable skillset going forward." - Alex Krentsel (21:07)"Part of the challenge is also deciding what to build, now that it's become really easy to actually do the building." - Alex Krentsel (23:08)FollowSpotify https://podcasters.spotify.com/pod/show/1st10podcast Amazon https://music.amazon.com/podcasts/7e8ec9af-f38c-4cd9-8c68-1c1dd4516b 27/1st10-podcastApple: https://podcasts.apple.com/us/podcast/1st10-podcast/id1760411207 Podcast: https://www.1st10.com/podcast YouTube: https://www.youtube.com/@1st10podcast Links:Alex Krentsel on LinkedIn: https://www.linkedin.com/in/alex-krentsel/Alex’s Website: https://www.krentsel.comRecent paper co-authored by Alex and Ion Stioca: https://arxiv.org/abs/2609.12039. It presents how we are thinking about the gaps in agentic systems (or agent-created systems). Bach, Partita II, Allemande, performed by Alex Krentsel: https://www.youtube.com/watch?v=8bxF6tpL2doBach - Violin Partita no. 2 in D minor BWV 1004 - Sato: https://www.youtube.com/watch?v=44Wz92zQe04Bach - Chaconne from Violin partita no. 2 in D minor BWV 1004:https://www.youtube.com/watch?v=7y4lcQ7BTLwMusic by Roman Senyk from PixabayProducer: Shrikant Joshi (https://42quirks.com)

What comes after ChatGPT and Claude and Gemini?A founder who sold an AI company to Microsoft for $160 million just told us that LLMs are heading toward a dead end. And he's betting his next company on what comes after.On this episode of the 1st10 Podcast, Sam Pasupalak, Co-Founder & CEO of Skyfall AI, sits down with host Boris Epstein to discuss what a post-LLM world looks like. Sam believes LLMs will become the low-level execution layer underneath something much more powerful: Enterprise World Models. In this conversation, Sam explains:The two research breakthroughs that made ChatGPT inevitableA simple analogy that exposes why LLMs are fundamentally limitedWhat world models are and how they differ technologically from LLMsWhat a business simulation video game revealed about LLMs than no benchmark couldWhy Sam thinks humans should eventually be freed from operational workHow Skyfall plans to acquire and autonomously operate businessesWhy an AI CEO is much harder to build than an AI software engineerandWhy Sam is hiring hungry researchers for Skyfall and not established AI starsThe thing that WILL stick with you is how calmly Sam walks you through what it actually looks like to hand the keys of a real business to an algorithm. Like, someone who's already seen the next wave forming and is just waiting for the rest of the industry to catch up.Chapters00:00 Preview of the episode01:00 Introductions03:23 Building AI Before AI Was Cool06:37 Did Sam Predict Today's AI Boom?10:01 Was AI Just Another Bubble?13:46 Are We at Peak LLM?17:53 "Smart Soccer Analyst"21:32 What Is an Enterprise World Model?26:06 The Beginning of the End for LLMs28:48 Skyfall's Bet: The Autonomous Enterprise34:33 Putting Money Where (World) Models Are38:45 What Happens to the Human CEO?41:40 Hiring for Hunger, not Hype44:58 Bold Prediction: AI in Politics by 2050?47:14 The Future of Autonomous EnterprisesQuotes"When I sold my company to Microsoft, I was like, 'I'm never gonna do a technology-first company again.' I promised myself that." - Sam Pasupalak (10:54)"LLMs are like a very smart soccer analyst - goes through all the videos about soccer on the internet, reads about everything about soccer on the internet, but has never ever played a game of soccer in real life." - Sam Pasupalak (17:53)"The Scaling Laws moment for world modeling is coming up in the next 12 to 18 months." - Sam Pasupalak (26:41)"Let's say you throw trillions of dollars, increase the context window as much as possible, it will not matter. It doesn't matter." - Sam Pasupalak (28:38)"We ran the best LLMs, and we found that LLMs just can never run a theme park. It's impossible." - Sam Pasupalak (36:48)"If you want to build a completely autonomous enterprise, you have to buy the enterprise. You have to run all the algorithms across all the systems within the enterprise." - Sam Pasupalak (37:33)FollowSpotify: https://podcasters.spotify.com/pod/show/1st10podcast Amazon: https://music.amazon.com/podcasts/7e8ec9af-f38c-4cd9-8c68-1c1dd4516b 27/1st10-podcastApple: https://podcasts.apple.com/us/podcast/1st10-podcast/id1760411207 Podcast: https://www.1st10.com/podcast RSS: https://anchor.fm/s/f951319c/podcast/rss YouTube: https://www.youtube.com/@1st10podcast Website: www.1st10.comPodcast: www.1st10.com/podcast Twitter www.x.com/1st10engineersLinkedIn: www.linkedin.com/company/1st10/LinksSam Pasupalak on LinkedIn: https://www.linkedin.com/in/sam-pasupalak-57b4a220aSkyfall AI: https://www.skyfall.ai/ (LinkedIn: https://www.linkedin.com/company/skyfallai)Sam’s OpenReview Page: https://openreview.net/profile?id=~Sam_Pasupalak1Maluuba, on Wikipedia: https://en.wikipedia.org/wiki/MaluubaMusic by Roman Senyk from PixabayProducer: Shrikant Joshi (https://42quirks.com)