Drew Breunig

*slide whisle*

using claude code at a standing desk feels like DJing

If you're doing applied AI research (especially system design, benchmarks, evals, efficiency, or ops) you should be submitting to the Conference on AI and Agentic Systems... caisconf.org/pages/cfp/

I guess we’re gonna end up building a coalition around the constitution. Full circle.

The joke is that we developed LLMs to handle fuzzy problems, then LLMs invent a new language because natural language isn’t precise enough, so we’re back to code.

Anthropic comms is crazy.

“Here’s a 38-page essay about the existential threat of AI and how underprepared we are.”

*30 min later*

“Also you can now connect your B2B SaaS apps to Claude.”

a meme metal, supported by a fig leaf of some underlying truth.

For people wondering how they can price so high (and why many advertisers will pay for it) read this: www.dbreunig.com/2022/06/06/w...

They’re in the “shiny object phase”.

Why local news matters, in a page.

Probably worth revisiting this bit I wrote in 2016… www.dbreunig.com/2016/08/04/a...

I wanted several files written to my machine.

I wanted to download a file from a Claude Code Web instance, not commit it to Github.

I don't know if you've tried this, but the first working solution it arrived at was outputting a gz file as base64 then asking me to paste it into a txt file.

Free gas promotion for video game goes awry in UK Automakers and car dealerships have used the old "Free Gas" promotion to good effect, so it's not surprising that companies outside the industry might try it, too. Electronic Arts gave it a go recentl...

Found it. They did it in LA and London. It went as you would expect. www.autoblog.com/features/fre...

EA was a client when I worked at an ad agency. At one point there was a promo where characters took over a gas station and gave away free gas.

Regarding the Claude “Soul Document”: I wonder if training models on documents like these influences our perception of these models as sentient.

In many ways, LLMs are playback machines, which we talk to as human. It reminds me of facilitated communication. en.wikipedia.org/wiki/Facilit...

Dara visualization in the wild.

Job applicants are winning the AI arms race against recruiters Companies have been hit by a wave of chatbot-generated applications

An irony of the AI age is that relationships matter more. www.economist.com/business/202...

GitHub - dbreunig/just-bash-py: Python Bash emulation for agents, a port of vercel-labs/just-bash Python Bash emulation for agents, a port of vercel-labs/just-bash - dbreunig/just-bash-py

Got nerd-sniped into shepherding Claude while it implemented just-bash in Python.

It's a good continuation from whenwords. The test coverage is high (the orig repo and Bash itself).

I really like this idea of emulation as an option prior to using a full sandbox. github.com/dbreunig/jus...

AI writing feedback always feels a little off to me… Like it's been post-trained to death on linear reasoning sequences, or something.

You can bet $10k on the Dodgers making the playoffs, to win $100.

Or you can put $10k in a 1 year CD at 4% and get $400.

It is amazing how poorly labeled and undersold that chart is given the importance of what it conveys.

The gap between the two lines is the value of a human in the loop. www.anthropic.com/research/ant...

Terence Tao's takes on how AI helps mathematicians resonate the most with my feeling about coding agents as well.

If I wanted to start a woo cult, its prime obsession would be the near perfect overlap of the moon and sun during eclipses. Constantly surprised no one is running with this.

Whenwords Limitations · Issue #3 · dbreunig/whenwords Critical Limitations and Conceptual Concerns This library does not take into account Unix timestamps. This raises a more fundamental question: what does it actually mean to be relative? Is relativi...

You know it's a good Monday when you get to quote Dr. Dan Streetmentioner's "Time Traveler's Handbook of 1,001 Tenses" on a Github issue before noon. github.com/dbreunig/whe...

YouTube video by Databricks Let the LLM Write the Prompts: An Intro to DSPy in Compound AI Pipelines

Here's my intro talk on it: www.youtube.com/watch?v=I9Zt...

And the write up, if you prefer: www.dbreunig.com/2025/06/10/l...

Your last two responses are basically the pitch in a nutshell and goal of the framework. To properly engineer the squishing LLM bits.

The maintainability of writing structured specs, not natural language (in addition to the optimization infra) is such a win for the exact scenario you're describing.

Would love to chat with you about DSPy and this use case.

How to Fix Your Context 6 tactics for fixing your context and shipping better agents. As Karpathy says, building LLM-powered apps means learning to ‘pack the context windows just right’—smartly deploying tools, managing info...

Yes. It's close to the "Tool Loadout" context fix I wrote about awhile back. Have been tracking a few implementations. www.dbreunig.com/2025/06/26/h...

YouTube video by E4 How to make photos look vintage- helpful tutorial from Sam Campbell | Remote Comedy from The Paddock

This is a perfect but of humor for a very specific niche of shareware enthusiast. cc @ernie.tedium.co youtu.be/z34DIbpu46M

This is performance art, as much as anything else. The dumbest post Ive written in a bit, but the one that raises the most questions.