attogram

More: https://github.com/attogram/dada-vibing/issues/184

Ahh I see. I'm doing it per repo. This is distributed.

I once tried to teach an agent to have a 'pain reflex' and save memory. The experiment failed horribly.

Interesting. I have hypothesis that markdown in repo is better.

See https://github.com/attogram/academic-vibing/blob/main/academ...

Summary? Is it "markdown as memory"?

8b max on a std 16gb macbook. Anything more and your mac is toast

"Agents MUST do x" never works. 100% fail rate.

And sanitizing via the LLM never really works.

Yes, the repo name has two n’s (`phpcoinn`) - phpcoin was already taken, so they added an extra n. Not a typo :)

I recently found PHPCoin, a blockchain written entirely in PHP. A niche project!

Repo: https://github.com/phpcoinn/node

Website: https://phpcoin.net

- A real, live L1 chain, with a small but active dev team

- Supports smart contracts, dapps, and web apps

- Readable and hackable PHP code

"It works. That's annoying." Indeed!

Would be cooler if support for local llms was added. Currently only has support for anthropic and openai. https://github.com/samrolken/nokode/blob/main/src/config/ind...

Good catch!

For folks curious about the rm -rf see https://github.com/attogram/bash-screensavers/blob/main/gall... line 339

Agreed! Known bug that will get squashed...

Some artifact from asciinema maybe. Only shows up in the preview gifs. Needs to be fixed!

See the spread-the-word script in https://github.com/attogram/bash-screensavers/tree/main/spot... This script generates a series of shell commands to create a "spotlight" message on the main GitHub repository page. It does this by generating commands that make trivial changes to the top 12 files and directories and then commit those changes with custom messages.

Gallery of current screensavers: https://github.com/attogram/bash-screensavers/blob/main/gall...

First, why no support for local models?

Second, the emojis in the readme scream AI generated more than any emdash ever could.

Exactly. Give us a better URL with the full content, otherwise this post is useless.

Sorry, but I see nothing about parasitic AIs here. Just lots of stuff about prompt engineering combined with unstable humans.

Does USDC make profit in same range as USDT?

So you are worried that you will be assimilated by the Microsoft borg. Valid concern. But all that free stuff is hard to stay away from!

The "Boundless Care" concept seems akin to Hinkins' Maternal AI.

That's the press release. For better write up see the register: https://www.theregister.com/2025/09/05/us_army_enlists_ai_st...

"Because handing battlefield ID to an algorithm has never gone wrong before, right?"

If I have to pay and still get ads, then no. Sorry.

Interesting progression of the vibe coding meme, from the original defintion in feb 2025 by Andrej Karpathy, to today

Nice. Suggest you add local llm options: Ollama, vLLM, LM studio, even llama.cpp

So, in summary: a lot of startup founders do not play well with others, often due to psychological issues (ADHD, etc).

Great paper title.

TLDR: "We argue that data systems need to adapt to more natively support agentic workloads."

Got a lot of manual steps to publish. It all could be done automatically on commits and prs with proper workflow config.

Url is paywalled.

And congrats Mistral!