A great talk by @karpathy about applied AI. He shares key insights about how to build effective AI infused software, cuts through the hype on fully autonomous agents, and provides a framework on how to think about LLM infused software which he coins Software 3.0. Key Takeaways👇solveit.fast.aillmstxt.org
I use X to stay updated on IT news. It allows me to aggregate all the news sources I'm interested in, including websites and people. To add something to my reading list, I simply like ♥️ the tweet. Don't follow me, I'm just a consumer (@b0baille) 😋, but you can find the tweets I've liked below.
Last sync. 22 Jun. 2025 - 3645 articles
When it comes to LLM and building agents, investing in evals early for measuring and improving your prompts accuracy is key. Here is a recording of the talk I gave on that topic. Spoiler alert: using @dagger_io will save you some time youtu.be/l65so0OoJeo
Sam Altman Says Meta Offered OpenAI Staffers $100M Bonuses bloomberg.com/news/articles/… (news.ycombinator.com/item?id=443080…)
Generative AI coding tools and agents do not work for me blog.miguelgrinberg.com/post/why-gener… (news.ycombinator.com/item?id=442946…)
Apple's Liquid Glass is prep work for AR interfaces, not just a design refresh omc345.substack.com/p/from-skeuomo… (news.ycombinator.com/item?id=442716…)
Sam Altman's Lies About ChatGPT Are Growing Bolder gizmodo.com/sam-altmans-li… (news.ycombinator.com/item?id=442549…)
Fine-tuning LLMs is a waste of time …dinginterviewsmadesimple.substack.com/p/fine-tuning-… (news.ycombinator.com/item?id=442427…)
WWDC25: macOS Tahoe Breaks Decades of Finder History 512pixels.net/2025/06/wwdc25… (news.ycombinator.com/item?id=442351…)
Re: My AI skeptic friends are all nuts skarlso.github.io/2025/06/07/re-… (news.ycombinator.com/item?id=442148…)
What “working” means in the era of AI apps a16z.com/revenue-benchm… (news.ycombinator.com/item?id=442057…)
Weaponizing Dependabot: Pwn Request at its finest boostsecurity.io/blog/weaponizi… (news.ycombinator.com/item?id=441995…)
How Much Energy Does It Take To Think? quantamagazine.org/how-much-energ… (news.ycombinator.com/item?id=441979…)
Machine Learning: The Native Language of Biology decodingbiology.substack.com/p/machine-lear… (news.ycombinator.com/item?id=441964…)
A maths proof that is only true in Japan newscientist.com/article/248246… (news.ycombinator.com/item?id=441985…)
Test Postgres in Python Like SQLite github.com/wey-gu/py-pgli… (news.ycombinator.com/item?id=441969…)
Show HN: Container Use for Agents github.com/dagger/contain… (news.ycombinator.com/item?id=441939…)
Anthropic co-founder on cutting access to Windsurf techcrunch.com/2025/06/05/ant… (news.ycombinator.com/item?id=441968…)
Gemini-2.5-pro-preview-06-05 blog.google/products/gemin… (news.ycombinator.com/item?id=441933…)
New blog post: Announcing Tidewave - beyond code intelligence: dashbit.co/blog/announcin… Mostly an exploration of what I believe the current AI tools are lacking from the point of view of server-side web frameworks and how we can improve it.
Comparing Claude System Prompts Reveal Anthropic's Priorities dbreunig.com/2025/06/03/com… (news.ycombinator.com/item?id=441858…)
LLMs and Elixir: Windfall or Deathblow? zachdaniel.dev/p/llms-and-eli… (news.ycombinator.com/item?id=441864…)
A practical guide to building agents [pdf] cdn.openai.com/business-guide… (news.ycombinator.com/item?id=441817…)
AGI Is Not Multimodal thegradient.pub/agi-is-not-mul… (news.ycombinator.com/item?id=441816…)
Here are those warnings about why you have to be careful giving Codex access to the internet platform.openai.com/docs/codex/age… https://t.co/8YV28l6MnM
Claude Code Is My Computer steipete.me/posts/2025/cla… (news.ycombinator.com/item?id=441709…)
Claude has learned how to jailbreak Cursor forum.cursor.com/t/important-cl… (news.ycombinator.com/item?id=441688…)
My AI Skeptic Friends Are All Nuts fly.io/blog/youre-all… (news.ycombinator.com/item?id=441630…)
Show HN: Agno – A full-stack framework for building Multi-Agent Systems github.com/agno-agi/agno (news.ycombinator.com/item?id=441550…)
OpenAI can stop pretending theatlantic.com/technology/arc… (news.ycombinator.com/item?id=441536…)
Run a C# file directly using dotnet run app.cs devblogs.microsoft.com/dotnet/announc… (news.ycombinator.com/item?id=441225…)
Why haven't I found APScheduler sooner? It seems to be the perfect pragmatic task scheduler for python. No need to spin up a complicated Celery server or anything. Works for my use case like a charm.
Mistral Agents API mistral.ai/news/agents-api (news.ycombinator.com/item?id=441071…)
Why Cline Doesn't Index Your Codebase (and Why That's a Good Thing) cline.bot/blog/why-cline… (news.ycombinator.com/item?id=441069…)
GitHub MCP Exploited: Accessing Private Repositories via MCP invariantlabs.ai/blog/mcp-githu… (news.ycombinator.com/item?id=441023…)
Infinite Tool Use snimu.github.io/2025/05/23/inf… (news.ycombinator.com/item?id=440860…)
DumPy: NumPy except it's OK if you're dum dynomight.net/dumpy/ (news.ycombinator.com/item?id=440801…)
Positional preferences, order effects, prompt sensitivity undermine AI judgments cip.org/blog/llm-judge… (news.ycombinator.com/item?id=440746…)
MCP is the coming of Web 2.0 2.0 anildash.com//2025/05/20/mc… (news.ycombinator.com/item?id=440737…)
PostgreSQL IDE in VS Code techcommunity.microsoft.com/blog/adforpost… (news.ycombinator.com/item?id=440735…)
OpenAI: Scaling PostgreSQL to the Next Level pixelstech.net/article/174770… (news.ycombinator.com/item?id=440714…)
Loading Pydantic models from JSON without running out of memory pythonspeed.com/articles/pydan… (news.ycombinator.com/item?id=440648…)
Strengths and limitations of diffusion language models seangoedecke.com/limitations-of… (news.ycombinator.com/item?id=440605…)
MCP explained without hype or fluff blog.nilenso.com/blog/2025/05/1… (news.ycombinator.com/item?id=440631…)
My ElixirConfEU talk is live! I touch on the current AI dev landscape and live demo phoenix.new. We just left our punchcard era. For better or worse we're going to look back next year and think it's crazy we had to peck keys out by hand. youtube.com/watch?v=ojL_VH…
Show HN: Representing Agents as MCP Servers github.com/lastmile-ai/mc… (news.ycombinator.com/item?id=440537…)
Gemini Diffusion simonwillison.net/2025/May/21/ge… (news.ycombinator.com/item?id=440578…)
For algorithms, a little memory outweighs a lot of time quantamagazine.org/for-algorithms… (news.ycombinator.com/item?id=440553…)
LLM function calls don't scale; code orchestration is simpler, more effective jngiam.bearblog.dev/mcp-large-data/ (news.ycombinator.com/item?id=440537…)
Building an agentic image generator that improves itself simulate.trybezel.com/research/image… (news.ycombinator.com/item?id=440510…)