utopian tech bro, crunchin numbers. Research Engineer @PAM/HarmixAI
Madrid/London
Joined March 2015
- Tweets70
- Following345
- Followers43
- Likes127
Vitalii Ratushnyi retweeted
The context a long-running agent works over across days easily runs to millions of tokens and the field still has no consensus on what memory should even look like.
@Misreadable is the Lead Research Engineer at HarmixAI working on long-term memory for agents with inputs of 1–10 billion tokens, and a former quant, now based in Madrid.
He'll compress two years of research and production experience into practical recipes for scalable agent memory.
I generated 200 landing pages with both Claude Code and Codex, and they are so look alike. The models are too noisy and generate lots of text (as a consequence of CoT), and for design patterns, it's impossible to get a great landing page without additional prompting/skills.
Spent 2 weeks in Rio back in April:
- Use only Uber Black (which will be an average european Uber X car);
- Keep your windows shut during the ride;
- Do not keep valuable belongings easy to retrieve from the window during the ride (better keep it in between seats somewhere in legspace;
- Leblon best place to stop;
- Barra da Tijuca is amazing surfing and more safe;
- Do not sleep on the beach and keep eye on cheap stuff sellers;
- Use SPF;
Pao de queijo is amazing out there!
looking for real VC deals/term sheets/SAFEs for vcplay.fyi
Looking for term sheets/SAFE's to gather up real numbers. vcplay.fyi/
No need for a career transition grant, but highly bullish on AI control/Oversight for Agentic attacks against other agents (sales/operations/support).
"I hate idea of a hosted second brain but also don’t wanna be my own It department."
You spotted the issue quite accurately here. I don't think labs will make it ideal as there is a lots of free energy in pulling context from your integrations, and labs will be limited by compute.
Figured I spent lot of compute on deep research and bet many of you are doing the same. Decided to assemble results in a structured way in the repo, both for human/agent readers. Only important tech and a structured format for easier context navigation. Pls contribute with ideas and star the repo: github.com/MisreadableMind/a…
