Claude Code may end up becoming the personal agentic framework that I've been looking for.
In the past when I needed some agentic behaviour (example: answer questions based on some documents), I used to look at hand rolling a solution. It's time consuming and not particularly easy to build and …
The latest LiteLLM package has been compromised. This is an issue because LiteLLM is used as a third party dependency for many python AI tools.
If you installed any python AI tool recently and it pulled in the latest litellm as a transitive dependency then you are at risk. You …
Whoa, whoa, whoa. The darling of the Python ecosystem, the team that brought us ruff, uv, ty and pyx and forever changed the python tooling landscape is now joining OpenAI.
Pretty sure all these tools will be deeply integrated into Codex, which is good news for python developers.
Also a …
Sometimes old is gold.
One of the key aspects of agentic coding is to have a good validation harness in place. Now, I've really seen three levels that developers go through when it comes to validation harness:
First, many devs simply not have a validation harness. You ask the agent …
Poor Perplexity. Everyone in our household uses it through my paid account.
Perplexity implements agentic memory over past conversations to improve results and it is constantly confused. So funny when I ask a deep research question on something highly technical and it comes back saying "Since you are studying in …
MCP or CLI? There was a huge amount of hype over MCP around a year ago, but today it seems like the command line interface has quietly won the game.
That's right, it you want your agent to interact with an external system, building a CLI interface seems the way …
There has been a lot of talk recently about OpenClaw, Moltbook and "sentient" (?) behaviour. So this is a good time to plug one of my favourite books ever.
I first came across "Turtles, Termites, and Traffic Jams: Explorations in Massively Parallel Microworlds" around 15 years ago when I borrowed it …
Developers who are worried that code produced by AI isn't always 100% accurate need to remember that humans aren't 100% accurate either.
That's why we do code reviews, linting, unit testing, integration testing, manual testing, continuous integration, demos... There is a whole suite of additional activities around coding to take …
"Hire a good talent, then get out of the way and give them room to do their job". (ie, don't meddle/micromanage) Is this good advice?
Paul Graham's latest article on Founder mode is making waves. In the article, Paul talks about the common management wisdom that I've quoted above …
For an LLM to work, it needs to somehow assign a meaning to a word. The problem is words can have different meanings in different contexts and be related to different words in multiple ways.
For example, Python and Java are related, so is Python and Cobra, and so is …