I'm primarily using my @fullstackpython handle now, follow me over there instead.

Last week 400 AI Engineers got together for the @aiDotEngineer Summit to talk about building agents. Here's some photos, stories, and takeaways. Many thanks to @swyx and @Benghamine for bringing the show to NYC!

🔥 Unlock the power of DeepSeek-R1 🔥 Get started with the open-weights distilled DeepSeek-R1 model via @OpenRouterAI and take full control of your prompts with #LaunchDarkly—all in just a few quick steps.

Super easy way to improve the effectiveness of coding models: First, take your prompt and add "Don't write the code yet — just write a fantastic, detailed implementation spec." Then, after the AI responds, say "Now, implement this perfectly." Makes a huge difference.

🤯 Achieving 5.7x reduction in costs using local and cloud models together while maintaining 97.9% of cloud model performance. @HazyResearch's Stanford team is showing us what's possible. @Avanika15 @dan_biderman @EyubogluSabri along with @avnermay @scott_linderman @james_y_zou

People ask me a lot: what advice do you have to build a 6 figure side hustle? What they want to hear: - learn to code with AI - here’s how to monetize - easy ways to acquire users What is real: - have 2 decades of credibility in a space - spend 5 years helping niche audience -

Using my favorite product (@LaunchDarkly AI configs) to test whether 4o or o3-mini writes better docs. Right now, o3 writes *more* and takes about 45% more time to do so. TBD if users like them better (need to hook up 👍👎)

"This is really good software. This small team at Apple appears to be almost single-handedly giving NVIDIA’s CUDA a run for their money!" Kudos to @awnihannun and the whole MLX team from @simonw /cc @tim_cook

It's ridiculous how much LLMs hallucinate - whenever I read 60 million books I remember every detail perfectly.

levels in tech: -junior eng -mid level eng -senior eng -staff eng -senior staff eng -principal eng -farmer

🚀 Drop @deepseek_ai r1 into your Python app—fast & hassle-free. With #LaunchDarkly AI Configs (early access!), integrate new models instantly. No setup, no delays—just pure AI innovation. ⚡ See it in action:

Watch Cursor control my browser. It can see what's happening, including network and console logs. Now my Cursor can debug and fix web site problems without any copy-pasta. HUGE speed up for web dev - with this Cursor can compete with v0 and replit at autonomous web app dev.

My Flask Mega-Tutorial ebook is now also available from The Pragmatic Bookshelf store, without DRM! I'm super excited to offer this additional channel for those who do not want to be locked in to Kindle. https://t.co/THwTh8q4t4

if openai's operator can navigate the aws console autonomously i'll know agi has been achieved

Been playing a lot with our new @LaunchDarkly AI config product and a meta learning is that a nice chatbot is a yapper and therefore consumes more tokens. Save costs; build mean AI.

You should be able to set your title and emblem in corporate meetings, a la mw2. Unlocking them as you do things at the company. Take down prod? Nuke emblem. Survived a reorg? Juggernaut emblem.

“How did you know i was a nextjs developer?” my living room:

GS: Only 6% of US companies report using AI

Discovered a very interesting thing about DeepSeek-R1 and all reasoning models: The wrong answers are much longer while the correct answers are much shorter. Even on the same question, when we re-run the model, it sometimes produces a short (usually correct) answer or a wrong