Ege Altan

co-founder & ceo at @GoodeyeLabs
prev. meta. phd in biological and artificial neurons

[Ege Altan @egealtan]

Here's why data scientists are making a comeback:
https://t.co/kSE8sv4qQ7

[Ege Altan @egealtan]

@sdianahu very true. wrote about the connection between evals, taste and experimentation a while back:

https://t.co/kSE8sv3T0z

[Ege Altan @egealtan]

@pauliusztin_ Start with #8 - error analysis. the most important step, by far

[Ege Altan @egealtan]

@lateinteraction agreed. we build and optimize harnesses because models by themselves are disappointing for most use cases that people care about

[Sebastian Raschka @rasbt]

PS: I finally got around to trying out @randal_olson 's Tufte Test tool to prettify the benchmark plot. Great tool 👌!
https://t.co/146ZnHsbgn

[Ege Altan @egealtan]

hide yo kids, hide you claws

[Boris Cherny @bcherny]

Mythos is very powerful, and should feel terrifying. I am proud of our approach to responsibly preview it with cyber defenders, rather than generally releasing it into the wild.

Model card here: https://t.co/HjhknJcRKQ

[Ege Altan @egealtan]

@tunguz did you say make no mistakes?

[Ege Altan @egealtan]

is there anything cleaner than git to manage agent skills across claude code/cowork/desktop? across cursor and claude?

[Ege Altan @egealtan]

@justinskycak also true for working with agents

[Ege Altan @egealtan]

@Stone_Tao evals

[Ege Altan @egealtan]

@Yuchenj_UW LLM output quality is heavily correlated with input quality

[Ege Altan @egealtan]

@RayFernando1337 interesting. the best ones I know are strategy gamers, but not necessarily APM heavy ones. exactly 0 of them are active gamers though. they are too busy playing with their agents now

[Ege Altan @egealtan]

@davidfowl yes but don't do this (meme from @HamelHusain ):

[Ege Altan @egealtan]

@sh_reya yes! i'm something of a token juggler myself

[Kaps 🇦🇪 | AI • Tech • Life @caprikaps]

Been seeing this everywhere. the smartest people in my circle now see every problem as an AI problem. every conversation comes back to what claude or gpt said. every decision gets validated through a model first. at some point the tool became the lens and nobody noticed when they stopped thinking independently. that's the psychosis - not that AI is bad, but that the smartest people stopped trusting their own judgment over a model's confidence

[Ege Altan @egealtan]

"proficiency in agent orchestration" is the new "proficiency in microsoft office"

[Ege Altan @egealtan]

ship to reduce entropy

[Ege Altan @egealtan]

@lateinteraction surge pricing kicks in when your agent starts editing files it shouldn't

[Omar Khattab @lateinteraction]

why isn’t there a human harness gig economy, like, what’s the Uber of babysitting your claude code runs

[Ege Altan @egealtan]

@Turkcell bu the hound degil mi ya

[Ege Altan @egealtan]

don't let your agents go npm install trust-me-bro without sandboxes, approvals, audit trails

[Feross @feross]

🚨 CRITICAL: Active supply chain attack on axios -- one of npm's most depended-on packages.

The latest [email protected] now pulls in [email protected], a package that did not exist before today. This is a live compromise.

This is textbook supply chain installer malware. axios has 100M+ weekly downloads. Every npm install pulling the latest version is potentially compromised right now.

Socket AI analysis confirms this is malware. plain-crypto-js is an obfuscated dropper/loader that:

• Deobfuscates embedded payloads and operational strings at runtime
• Dynamically loads fs, os, and execSync to evade static analysis
• Executes decoded shell commands
• Stages and copies payload files into OS temp and Windows ProgramData directories
• Deletes and renames artifacts post-execution to destroy forensic evidence

If you use axios, pin your version immediately and audit your lockfiles. Do not upgrade.