And the first episode of our ai/baking series is out! Move Fast & Bake Things.. where we bake cupcakes while talking about AI Kicking it off with my old friend @rickyrobinett. One of the real OGs of DevEx and now running developer relations at Google Cloud. Before that he ran developer community at Cloudflare and before that Twilio for about a decade. We've known each other since I was a high schooler going to hackathons. I threw together a Gemini recipe for cupcakes (of course) and we talked about the rebirth of hacker culture why his 9-year-old went from Cursor to Vim, and what it actually means to be a builder now. 0:00 – Intro & the Gemini cupcake recipe 3:02 – The hackathon boom and what's changed in 10 years 5:37 – Ricky's daughter went from Cursor to Vim at age 7 12:00 – What is a developer anymore? 13:15 – The irreducible kernel of building 19:06 – The halting problem as a weapon against your boss 22:10 – Anti-gravity browser testing 30:34 – How do you convince an AI skeptic? 35:38 – Are non-AI coders already left behind? 37:02 – From REST APIs to MCP 49:22 – Startup Bus throwback 53:52 – DevRel has always been the same two things 57:15 – Google AI Studio, Gemini CLI, and how Ricky actually works

Read our recap of the vibe coding olympics and how we organized the first event of it's kind and a dive into the winning projects! promptlayer.com/blog/we-host…

How do you stand out in the age of agents? Now that every website has cool animations. Now that every meal is a bowl. ==the antidote is soul== The best meals are made with love. The best garmentos obsess over button details. Good Design is taste. It conveys a message. It’s human. We’ve redesigned promptlayer.com and we’re really proud of the result. It’s inspired by my favorite restaurant in SF. Every icon is hand drawn.

Recap of the vibe coding olympics ~~ 🏆⛷️🍰 Round 1 (Qualifier): HTTP endpoint golf Round 2: Build something for bike delivery couriers Round 3: Re-imagine an AI-native Instagram Judgement criteria: vibes Round two winners were an app to bikers to calm down & relax. And also a dating app for bikers 😂 Final round winner was social media for abstract ideas & memories We'll have to do another soon! @RideJOCO @EndexAI @MLHacks @ExaAILabs @cursor_ai

The first ever Vibe Coding Olympics was a SUCCESS! Thanks to all who came & vibed Pure magic Special thanks @MLHacks @EndexAI @RideJOCO @cursor_ai @ExaAILabs

Can't wait for the vibe coding olympics on Thursday We just announced new prizes 🦞 First round will start 6:30p sharp in Soho Signup here vibeolympics.org/

PromptLayer MCP out now 🍰 A bunch of launches coming up... stay tuned

Feeling left out of the winter olympics? ☃️🎿 Not named Lindsey Vonn? Us too.. Presenting... "The Vibe Coding Olympics" 3 rounds. 1v1. All coding tools permitted. RSVP & apply to compete. Feb 26 NYC. vibeolympics.org/ by PromptLayer

LESSON: Get out of the model's way! When something doesn't work, the instinct is to add more. More guardrails. More tools. More structure. With LLMs, this instinct is often wrong. @andrewqu shared a great example through Vercel's text-to-SQL agent. He deleted 80% of their tools and gave the model access to bash and their raw semantic layer files. The agent got way way way better "We simplified our multi-phase, many-tool text-to-SQL agent down to a file system agent with bash. The simplification led to significantly improved performance and re-iterated the bitter lesson of not trying to fight against the model's ability.” ➡️ Every tool you add is a choice you are forcing the model to make. LESSON: Don’t Fighting Gravity We used to build complex DAGs to manage "probabilistic" models. No more. All that scaffolding was solving problems the model can now handle on its own. We assumed Claude would get lost in complex schemas, make bad joins, or hallucinate table names. In turn, we built guardrails for every turn. We were fighting gravity. We were forcing models into boxes they'd had outgrown. LESSON: Work Downhill, Not Uphill Models are trained on human data. It’s easier to steer into what they already know, rather than trying to teach them something new. If you're trying to teach a model a custom format or bespoke abstraction, you're working uphill. The model has way less training data on your special thing, so it's going to do a worse job. Work downhill instead. Standard languages beat custom DSLs. LLMs are amazing at writing Python, TypeScript, and SQL. Conversely, they have never heard of your proprietary query language. This is increasingly obvious for the top AI teams. They start with a bespoke XML language designed for agents. Then simplify to markdown results and tool calling. Now, most have simplified again and agents just write TypeScript directly. ➡️ Don't reinvent the wheel. If you can use Python or SQL, adapt your application to it, not vice versa. LESSON: Bash is The Universal Tool Call The answer is increasingly clear: give the model a file system and bash access. Both Claude Code and OpenAI's Codex use remarkably simple architectures. No complex DAGs. Just a two-layer while loop where the model can call bash commands. ➡️ The file system becomes your context manager. `grep`, `cat`, `ls` are 50 years old, included in loads of model training data, and still do exactly what we need. I often see Claude Code creating and running Python scripts solely as a way to test its own code. It writes a one-off Python file, runs it, checks the output, and moves on. No specialized "test runner" tool needed. ➡️ Code is the interface. Code is the handshake. Code is the ground truth. LESSON: AGI-Pilled Development Build assuming models will keep getting dramatically better. Today, you might need extra guardrails to ship something that works. But you should try to avoid it when you can, because every new model is still a step function improvement. Everyone still upgrades to the newest release without thinking twice. Start with the simplest possible architecture: model + file system + goal. Add complexity only when you've proven it's necessary. Often, the best thing you can do is get out of the model's way. Stop making choices for it. Stop constraining its reasoning. Stop building tools to protect it from complexity it can handle. ➡️➡️➡️ Behave like a human. Work downhill. Let the model cook.