Twitxy Saved

Javi™

Verified account This account is verified. Learn more

@javitm

CTO @theagilemonkeys.
Building agents with agents to spawn incredible things into existence.
Slopcore artist by accident.
Las Palmas de Gran Canaria theagilemonkeys.com Joined July 2008
1.1K Following 1.5K Followers 6.4K Posts

Latest posts 5

Pinned post

Javi™ Verified account @javitm
Attention is the most precious asset we have, especially when parallelizing work across several agents. How do you handle this without going crazy?
The Agile Monkeys Verified account @theagilemonkeys Nov 5
Agents multiply how much work we can run, not how much we can judge. Design every agentic workflow around the human attention it consumes. x.com/i/article/2118…
View quoted post on X
Javi™ Verified account @javitm
As Cypher said while eating the simulated steak in The Matrix: Ignorance is bliss.

OpenAI released 1,000+ security patches for the Linux kernel, the key software running banks, transportation, hospitals, and the rest of the internet.

Software wasn't safer last week; we just hadn't found the bugs yet, living with a false sense of security.

Take sshhole, patched this week. Under a specific combination of OpenSSH, musl Linux, and authentication settings, accounts meant to reject every password could accept one.

Unix used "*" to mean no password matches. A library used "*" to mean hashing failed. A password-length limit plus a timing-attack fix turned an error into a successful root login via a random string over 255 characters.

Several changes, made years apart, interacted in unintended ways.

Who could have figured? An AI agent did.

Welcome to the crazy security race of late 2026: use AI and every tool to find these bugs, or attackers will.
Javi™ Verified account @javitm
Researchers developed a method using AI and Lean (math verification software) to prove that quantum computers are better only when they can exploit certain structures in the data.
Today Physical Superintelligence PBC is announcing a proof of the Aaronson-Ambainis conjecture, one of the central open problems in quantum complexity, with major implications for quantum computing speedup.

psi.inc/research/aaron…

Thread 👇
View quoted post on X
Javi™ Verified account @javitm
Prompt engineering no longer makes sense. I do “context building” through conversations and references instead (not sure if there’s an official term). Once there's enough context in the session, the working prompt is simply “go ahead with the plan in ultracode mode”.