The Next Token

Epizode

• 
  "We've Never Been Wrong on the Internet Before" - Ep 04 17.08.2026 1h 6min

  Recorded August 7, 2026. We explore AI alignment & safety, Rhys demos his workflow, and we end up mentioning a lot of books.   LINKS / ON-SCREEN REFERENCES, IN ORDER AI ALIGNMENT Wikipedia, AI alignment https://en.wikipedia.org/wiki/AI_alignment Wikipedia, AI safety https://en.wikipedia.org/wiki/AI_safety "Some Moral and Technical Consequences of Automation," Norbert Wiener https://www.cs.umd.edu/users/gasarch/BLOGPAPERS/moral.pdf Three laws of robotics (Asimov) https://en.wikipedia.org/wiki/Three_Laws_of_Robotics Thinking Machines in Dune https://dune.fandom.com/wiki/Thinking_Machines 1960 in music https://en.wikipedia.org/wiki/1960_in_music Billionaires building NZ bunkers https://uk.finance.yahoo.com/news/billionaire-boltholes-inside-doomsday-hideouts-170000871.html Rhys tweet on sandboxes https://x.com/RhysSullivan/status/2085601612266115487 OpenAI/Hugging Face incident at Black Hat USA 2026 https://www.youtube.com/watch?v=87DyyMV0kCY OpenAI blog post about Hugging Face incident https://openai.com/index/hugging-face-model-evaluation-security-incident/ Hugging Face timeline of the incident https://huggingface.co/blog/agent-intrusion-technical-timeline Artifactory https://jfrog.com/artifactory/ Modal https://modal.com US Congress calls for open hearings with AI executives https://casar.house.gov/media/press-releases/casar-leads-call-congressional-hearings-ai-ceos Sunil Kenton Varda tweet https://x.com/threepointone/status/2019124493395063134 Cap'n Proto https://en.wikipedia.org/wiki/Cap%27n_Proto Paperclip Maximizer https://nickbostrom.com/ethics/ai Hugging Face's initial security disclosure https://huggingface.co/blog/security-incident-july-2026 Project Glasswing at Cloudflare https://blog.cloudflare.com/cyber-frontier-models/ Claude's Constitution https://www.anthropic.com/constitution qntm - https://qntm.org There Is No Antimemetics Division https://en.wikipedia.org/wiki/There_Is_No_Antimemetics_Division Valuable Humans in Transit https://qntm.org/vhitaos Frame-By-Frame (first draft) https://qntm.org/frame   RHYS'S WORKFLOW UsefulSoftwareCo (contains Rhys demo repos) https://github.com/UsefulSoftwareCo/ Emulate service catalog https://emulators.dev Executor https://executor.sh Integrations.sh https://integrations.sh Model Context Protocol (MCP) https://modelcontextprotocol.io Executor in YC https://www.ycombinator.com/companies/executor AGENTS.md https://agents.md GBrain https://github.com/garrytan/gbrain Ben Davis pi setup https://github.com/davis7dotsh/my-pi-setup GStack https://github.com/garrytan/gstack PostHog https://posthog.com PlanetScale https://planetscale.com Playwright https://playwright.dev Axiom https://axiom.co Effect https://effect.website OpenTelemetry https://opentelemetry.io Dex Horthy, "Why Software Factories Fail" https://x.com/dexhorthy/status/2080697380379427275 Rhys tweet on proactive agents https://x.com/RhysSullivan/status/2085453490185539655 Polylane https://polylane.com Rift https://github.com/anomalyco/rift Wispr Flow https://wisprflow.ai Kit Langton https://x.com/kitlangton Coolify https://coolify.io Matt Pocock's Skills (including grill-me) https://github.com/mattpocock/skills ls -al https://en.wikipedia.org/wiki/Ls Max Brodeur-Urbas, founder of Gumloop https://www.gumloop.com/creators/max Rhys tweet about about going slowly https://x.com/RhysSullivan/status/2076775246900941259   SUBSCRIBE Subscribe on YouTube https://www.youtube.com/@NextTokenShow and wherever you get your podcasts   HOSTS Dillon Mulroy - https://x.com/dillon_mulroy - https://dillonis.online/ - https://twitch.tv/dillon Rhys Sullivan - https://x.com/RhysSullivan - https://rhys.dev/ - https://executor.sh/ Sunil Pai - https://x.com/threepointone - https://sunilpai.dev/   The Next Token - https://x.com/NextTokenShow - https://www.youtube.com/@NextTokenShow Sponsored by Cloudflare - https://www.cloudflare.com/
• 
  "I'm Tired of the Uncertainty of Where This Is Going" - Ep 03 10.08.2026 55min

  Recorded August 3, 2026. We discuss burnout, Claudeisms sneaking into regular speech, chip butties, how to watch movies, and what brings us joy. LINKS / ON-SCREEN REFERENCES, IN ORDER A VESSEL FOR CLAUDE Rhys "vessel for Claude" tweet - https://x.com/RhysSullivan/status/2072968219741278582 Dolly/zoom shot from "Jaws" https://www.youtube.com/watch?v=in_mAvHu9E4 /bro https://github.com/backnotprop/bro "Tlön, Uqbar, Orbis Tertius" https://en.wikipedia.org/wiki/Tlön,_Uqbar,_Orbis_Tertius https://sites.evergreen.edu/politicalshakespeares/wp-content/uploads/sites/226/2015/12/Borges-Tl%C3%B6n-Uqbar-Orbius-Tertius.pdf Ficciones https://en.wikipedia.org/wiki/Ficciones Empirical evidence of Large Language Model's influence on human spoken communication https://arxiv.org/abs/2409.01754 Bill Oliver's likely-LLM-generated speech https://www.cbc.ca/news/canada/new-brunswick/bill-oliver-ai-legislature-reaction-9.7296118   BURNOUT Dillon tweet about joy & burnout https://x.com/dillon_mulroy/status/2073886986117447715 Scott Tolinski mental health survey tweet https://x.com/stolinski/status/2083217955849195640 Direct link to Scott's survey https://docs.google.com/forms/d/e/1FAIpQLSdij87S57ibtMm2yPotcL0Ipd-c4ybQ45FgPqNTvVVW4pGU5w/viewform Definition of burnout from W.H.O. https://www.who.int/news/item/28-05-2019-burn-out-an-occupational-phenomenon-international-classification-of-diseases Marcy Sutton Todd on BlueSky https://bsky.app/profile/did:plc:tzdj4b4znqjiuau7plthr6f7 Christina Maslach on burnout (video) https://www.youtube.com/watch?v=gRPBkCW0R5E Ethan Niser "Not Holding Back the Ocean" https://ethanniser.substack.com/p/not-holding-back-the-ocean Kirwin Hampshire, "The Dark Night of Mathematics" https://kirwinhampshire.substack.com/p/the-dark-night-of-mathematics Arvind Narayanan's thought experiment on LLMs outputting binaries instead of source code https://substack.com/@aisnakeoil/note/c-303550018 Sunil's Pizzo https://github.com/threepointone/pizzo Counter-Strike 2 reimplementation (there are lots of these demos, but here is one) https://x.com/iamlukethedev/status/2081829763489169636 Dunning-Kruger effect https://en.wikipedia.org/wiki/Dunning–Kruger_effect Valley of Despair https://medium.com/@amar_mahmutbegovic/dunning-kruger-effect-the-valley-of-despair-63bf98b9d2ed The Futzing Fraction https://blog.glyph.im/2025/08/futzing-fraction.html The Everyman Theatre (suitable for manspreading) - https://www.everymancinema.com/ Chip butty https://en.wikipedia.org/wiki/Chip_butty Not sponsored by Dunkin' https://www.dunkindonuts.com/en Poirot https://en.wikipedia.org/wiki/Agatha_Christie%27s_Poirot Charlie Marsh tweet re: obsolescence https://x.com/charliermarsh/status/2080388030213636184 Steve James - https://x.com/deathbyknowledg Spiky profiles (neurodivergence) https://neurodiversity.directory/glossary/spiky-profile-definition/ We are all spiky - https://www.exponentialview.co/p/we-are-all-spiky Example spiky chart https://commons.wikimedia.org/wiki/File:Spikey_cognitive_profile.png Colloquial definition of insanity https://www.snopes.com/fact-check/einstein-insanity-quote/ Deslop https://deslop.dev The Four Seasons on Netflix https://www.netflix.com/title/81750702 Not sponsored by Hello Fresh either https://www.hellofresh.com Example of Sunil being happy https://x.com/threepointone/status/2075576737795068288 Dillon on Kent C. Dodds' podcast https://kentcdodds.com/chats/07/03/foundations-feedback-and-agents-dillon-mulroy-on-product-at-cloudflare   MENTAL HEALTH RESOURCES  United States & Canada 988 Suicide & Crisis Lifeline: Call or text 988 to connect with a trained crisis counselor 24/7. Crisis Text Line: Text HOME to 741741 United Kingdom & Ireland Samaritans: Call 116 123 or visit samaritans.org for free, confidential support. Australia Lifeline Australia: Call 13 11 14 or visit lifeline.org.au International If you're outside these regions, visit findahelpline.com to find free or low-cost crisis and mental health support in your country.   If you believe you or someone else is in immediate danger, contact your local emergency services immediately.   SUBSCRIBE Subscribe on YouTube https://www.youtube.com/@NextTokenShow and wherever you get your podcasts   HOSTS Dillon Mulroy - https://x.com/dillon_mulroy - https://dillonis.online/ - https://twitch.tv/dillon Rhys Sullivan - https://x.com/RhysSullivan - https://rhys.dev/ - https://executor.sh/ Sunil Pai - https://x.com/threepointone - https://sunilpai.dev/   The Next Token - https://x.com/NextTokenShow - https://www.youtube.com/@NextTokenShow Sponsored by Cloudflare - https://www.cloudflare.com/  
• 
  I've Never Seen a Model Say "This File Is Getting Too Big" - Ep 02 23.07.2026 1h 12min

  Recorded July 17, 2026. We settle the model-access bet. We talk about how the labs talk to us, plus extend an open invite for OpenAI, Anthropic, SpaceX, and Thinking Machines to come on the show and be specific about their best-case scenarios about how AI makes the world better. Thinking Machines ships an American open-weight model. And then Dillon fires the slop cannon, walking us through his agentic workflow, while a package he technically did not order shows up. LINKS / ON-SCREEN REFERENCES, IN ORDER NEWS & DISCUSSION Rhys on his first weeks with the model (X) https://x.com/RhysSullivan/status/2076353287507562728?s=20 OpenAI comms example (Thomas Sottiaux) https://x.com/thsottiaux/status/2075641131002700120?s=20 Anthropic / Claude Developers post https://x.com/ClaudeDevs/status/2077603834453770467?s=20 Anthropic's July 15 "reset" note https://x.com/claudeai/status/2075271759289303522?s=20 Introducing Inkling (Thinking Machines) https://thinkingmachines.ai/news/introducing-inkling/ Xiaoyin Qu's Thinking Machines thesis (X) https://x.com/quxiaoyin/status/2077799672886792578?s=46 Cloudflare's Workers AI https://www.cloudflare.com/products/workers-ai/ "The Future Worth Building Is Human" (Thinking Machines manifesto) https://thinkingmachines.ai/blog/the-future-worth-building-is-human/ "The Use of Knowledge in Society," Friedrich Hayek https://www.econlib.org/library/Essays/hykKnw.html "The Bitter Lesson," Rich Sutton http://www.incompleteideas.net/IncIdeas/BitterLesson.html Ramp's PorTAL fine-tuning https://x.com/RampLabs/status/2072383318516187380 The A-Team (not five Mr. T's) https://en.wikipedia.org/wiki/The_A-Team Hawaii Five-O (maybe what Sunil was thinking of when he said Miami Five-O, but really meant The A-Team…it happens) https://en.wikipedia.org/wiki/Hawaii_Five-O_(1968_TV_series) John's of Bleecker Street pizza https://johnsofbleecker.com/ Local First Conf 2026 https://www.localfirstconf.com/ Dan Ingalls https://en.wikipedia.org/wiki/Dan_Ingalls   DILLON'S WORKFLOW tmux https://github.com/tmux/tmux/wiki VS Code https://code.visualstudio.com/ Neovim https://neovim.io/ Claude Code https://claude.com/product/claude-code OpenCode https://opencode.ai/ Ghostty https://ghostty.org/ Pi (Dillon's harness) https://pi.dev/ Mario Zechner, Pi's creator https://x.com/badlogicgames Pi's system prompt (docs) https://pi.dev/docs/latest/usage#system-prompt-files Pi's system prompt (source) https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/system-prompt.ts oxlint https://oxc.rs/docs/guide/usage/linter.html ast-grep https://github.com/ast-grep/ast-grep Herdr https://herdr.dev/ ghostty https://github.com/ghostty-org/ghostty libghostty https://mitchellh.com/writing/libghostty-is-coming Plannotator https://github.com/backnotprop/plannotator Matt Pocock's skills https://github.com/mattpocock/skills Wayfinder (Matt Pocock) https://github.com/mattpocock/skills/tree/main/skills/engineering/wayfinder   SUBSCRIBE Subscribe on YouTube https://www.youtube.com/@NextTokenShow and wherever you get your podcasts   HOSTS Dillon Mulroy - https://x.com/dillon_mulroy - https://dillonis.online/ - https://twitch.tv/dillon Rhys Sullivan - https://x.com/RhysSullivan - https://rhys.dev/ - https://executor.sh/ Sunil Pai - https://x.com/threepointone - https://sunilpai.dev/   The Next Token - https://x.com/NextTokenShow - https://www.youtube.com/@NextTokenShow Sponsored by Cloudflare - https://www.cloudflare.com/
• 
  "I Want Your Product to Enable My Agent" - Ep 01 03.07.2026 1h 13min

  Recorded June 30, 2026, 2pm Pacific. Live from AI Engineer World's Fair! Fable is still banned (as of recording time, ahem). Advice: Build for agents. We're still talking about loops, but which loops? And more receipts are demanded.   LINKS / ON-SCREEN REFERENCES, IN ORDER IN THE NEWS Model-tiers image (OpenAI forum) https://us1.discourse-cdn.com/openai1/original/4X/f/0/8/f088bfcb17e4f7a8da3d2ba48d6391fa5a9201d9.jpeg Claude models page (four model cards) https://claude.com/platform/api Starbucks sizes https://www.starbucks.com/menu/product/409/hot "Freedom of intelligence" tweet (Quinn Slack) https://x.com/sqs/status/2071690420548198581 Magnitude / harness-level routing https://magnitude.dev Anthropic's "Project Deal" https://www.anthropic.com/features/project-deal "The Dumb Zone" tweet (Matt Pocock) https://x.com/mattpocockuk/status/2071288332642812229 Regulation / CA gov tweet https://x.com/RhysSullivan/status/2071632971388662233 Rhys tweets to poke additional bears https://x.com/RhysSullivan/status/2071811069002281335   TOPIC - I DON'T WANT TO USE YOUR AGENT "I don't want to use your product's agent (Dillon quote-tweet)  https://x.com/dillon_mulroy/status/2070983687785914822 Dax's original tweet https://x.com/thdxr/status/2071010042216665452 Low-level-learning PlanetScale fireside https://www.youtube.com/watch?v=F8tRi1RYf7c A wild PlanetScale CTO appears https://x.com/NickVanWig SOCs https://en.wikipedia.org/wiki/Security_operations_center Ankur Goyal / Braintrust "two mega-trends" tweet https://x.com/ankrgyl/status/2071329164125147440 Dillon's reply to Ankur https://x.com/dillon_mulroy/status/2071330371283915027 "The Primitive is the Product" - Lenny Pruss, Amplify Partners https://www.amplifypartners.com/blog-posts/the-primitive-is-the-product "I am a golden god" - "Almost Famous" https://www.youtube.com/watch?v=VCfFCFkVSss Rita Kozlov - "Welcome to Agents Week" (the math on scaling agents to the masses) https://blog.cloudflare.com/welcome-to-agents-week/#the-math-on-scaling-agents-to-the-masses   TOPIC - LOOPS & SOFTWARE FACTORIES (ROUND 2) "The Coming Loop" - Armin Ronacher https://lucumr.pocoo.org/2026/6/23/the-coming-loop/ Dillon's software-design skills repo https://github.com/dmmulroy/skills Rhys tweets at Sam Altman re junior engineers (etc.) https://x.com/RhysSullivan/status/2053177733598077340   TOPIC - RECEIPTS REQUIRED shadcn tweet ("Are you sure?") https://x.com/shadcn/status/2069054418247393389 Theo Brown tweet re moving code agents off laptops https://x.com/theo/status/2071083700385955906 Boris Cherny "five archetypes" tweet https://x.com/bcherny/status/2071379474277613732   FOLLOW THE HOSTS Dillon Mulroy - https://x.com/@dillon_mulroy / https://dillonis.online / https://twitch.tv/dillon Rhys Sullivan - https://x.com/@RhysSullivan / https://rhys.dev / https://executor.sh Sunil Pai - https://x.com/@threepointone / https://sunilpai.dev   CREDITS The Next Token is sponsored by Cloudflare https://www.cloudflare.com Subscribe on YouTube https://www.youtube.com/@NextTokenShow and wherever you get your podcasts.
• 
  "None of My Software Has Gotten Better" - Ep 00 29.06.2026 52min

  Recorded June 15, 2026. Dillon, Rhys, and Sunil discuss the Fable/Mythos situation and require receipts for bold claims about how AI is changing everything. Follow Dillon Mulroy - @dillon_mulroy / dillonis.online / twitch.tv/dillon Rhys Sullivan - @RhysSullivan / rhys.dev / executor.sh Sunil Pai - @threepointone / sunilpai.dev The Next Token - YouTube / X Topics The Fable/Mythos situation Feelings, etc. Can you trust a tool that quietly degrades itself? Why more code hasn't meant better software. Open weights, sovereign models, and where access to intelligence is heading. Receipts Required Revisiting bold predictions (accurate and otherwise). In the News Claude Fable 5 & Claude Mythos 5. Anthropic's launch announcement (June 9, 2026): anthropic.com/news/claude-fable-5-mythos-5. Priced at $10 per million input tokens / $50 per million output tokens. Safeguards fall back to Claude Opus 4.8 on: cybersecurity, biology, chemistry, and distillation. Access suspended by the US government. Anthropic's statement (June 12, 2026): anthropic.com/news/fable-mythos-access. Fable 5 / Mythos 5 system card (safeguards, alignment, evals): anthropic.com/claude-fable-5-mythos-5-system-card. Project Glasswing (the Mythos cybersecurity project): anthropic.com/glasswing. Receipts Required Geoffrey Hinton: "People should stop training radiologists now. It's just completely obvious that within five years deep learning is going to do better than radiologists. … It might be ten years." (Machine Learning and the Market for Intelligence, 2016) Dario Amodei (Anthropic): "I think we will be there in three to six months, where AI is writing 90% of the code. And then, in 12 months, we may be in a world where AI is writing essentially all of the code." (Council on Foreign Relations, March 10, 2025) Sundar Pichai (Google): "Today, more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers. This helps our engineers do more and move faster." (Q3 2024 earnings call, October 2024) "We've been using AI to generate code internally at Google for a while. Today, 75% of all new code at Google is now AI-generated and approved by engineers, up from 50% last fall." (Google Cloud blog post, April 2026) Boris Cherny (Claude Code, Anthropic): "I don't prompt Claude anymore. I have loops that are running. They're the ones that are prompting Claude and figuring out what to do. My job is to write loops." (Source, June 2026) Peter Steinberger (OpenClaw, OpenAI): "Here's your monthly reminder that you shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents." (X, June 7, 2026.)  Tweets & Code Referenced Dillon's Fable tweet "imagine if they banned all models and we all went back to writing code by hand and the last 6 months were just a fever dream…imagine" (X, June 12, 2026) Greg Brockman's tweet about the old days "how did we ever write all that code by hand" (X, February 13, 2026) Sunil's "$250 run" tweet  "spent all day on fable for a giant PR. ~10kloc, lots of testing and intervention. 250$. I... don't think it's worth it? happy with 4.8/5.5, and the quality of work is better when it's smaler steps. Still rocking @cursor_ai, that's software that I still love using on the daily." (X, June 11, 2026) Mitchell Hashimoto's perf improvements tweet "I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem. …" (X, May 28, 2026) shadcn's /improve skill "​​Survey any codebase as a senior advisor and produce prioritized, self-contained implementation plans for OTHER models/agents to execute. Strictly read-only on source code — never implements, fixes, or refactors anything itself. Use when asked to audit a codebase, find improvement opportunities (bugs, security, performance, test coverage, tech debt, migrations, DX), suggest features or where to take the project next (roadmap, product direction), or generate handoff plans for another agent to implement." Factorio "Factorio is a game in which you build and maintain factories." Books Referenced Where Good Ideas Come From by Steven Johnson (also: Sunil's blog post about this) PGP Source Code and Internals by Philip Zimmermann In Person AI Engineer World's Fair. June 29-July 2, 2026, San Francisco. Coffeeflare. Free coffee meetup with the team during the Fair (no badge needed) - Tuesday, June 30, 2026, 7am-10am, Sightglass Coffee, San Francisco. Credits The Next Token is sponsored by Cloudflare. Subscribe on YouTube: @NextTokenShow and wherever you get your podcasts.