Seeing like an LLM
"I will run the tests again. I expect nothing. I am a leaf on the wind." an LLM while coding
"I will run the tests again. I expect nothing. I am a leaf on the wind." an LLM while coding
Consider all the steps between the total cost associated with one line of code written by an expert on a gigantic machine, to nowadays the cost of generating one line of code with an LLM
I'm sure people are starting to get a little tired of hearing about how AI solved this coding problem or that coding problem. If you look at X (urgh, I know...
I discovered in 2011 that I had won the genetic lottery: I am suffering from Retinis Pigmentosa. Since, my vision overall has been declining (worsening eyesight, narrowing field of view, loss of night vision), and my photophobia increased by a lot. The conditi…
In this blog post, we discuss the Model Context Protocol (MCP), prompt injection attacks, problems with LLM agency, and how to use MCP in production.
A bake off of Claude and Gemini in order to test their SQL performance against a ClickHouse database.
To make RL work on tasks we care about, we need better user simulators. What's missing?Co-written with Nick Tomlin.
Writing notes helps me think more clearly. I share them in case someone spots a gap in my thinking.
Code snippets are take from open-source tetragon, tracee and crictl projects.
Posted on 28 June 2025
AI agents are everywhere, but are they delivering real ROI? Most companies are stuck in an 'Intelligence Hyperinflation' trap, where disconnected AI tools yield dismal returns. The key to transforming that chaos into significant, measurable business value lie…
When you talk to an AI, you're talking to two other "people".
On StackExchange, someone asks why programmers talk about “calling” a function. Several possible allusions spring to mind: Calling a function is like calling on a friend — we go, we stay a while, we come back. Calling a function is like calling for a ser…
We conduct a randomized controlled trial (RCT) to understand how early-2025 AI tools affect the productivity of experienced open-source developers working on their own repositories. Surprisingly, we find that when developers use AI tools, they take 19% longer …
The JSON Programming Language. Contribute to W1LDN16H7/JPL development by creating an account on GitHub.
Hey everyone, After the last Cursor update, I’ve noticed the code suggestions have gotten way worse. Before, it used to generate clean, well-structured code that followed good practices. Now the code feels really basic, sloppy, and sometimes just wrong. For …
Today's AI automation is like using a robot to read your screen and click buttons. MCP-B gives AI direct access to your website's functions instead.