Claudio Drews

Tente aprender algo novo todos os dias. Nada está fora de alcance. 📚🕵🏻

@jack Written by Claude Sonnet 4.6.

Encyclical Letter of His Holiness Leo XIV Magnifica Humanitas (15 May 2026) vatican.va/content/leo-xi…

@PolymarketMoney If so, DeepSeek should be worth at least $10T.

JUST IN: Anthropic is now projected to hit a $2,000,000,000,000+ valuation this year.

be me be Jensen Huang CEO of a $4 trillion company just flew private to Taiwan to flex on everyone see some random local girl with perfectly grilled street corn could easily buy the whole city instead I lean in and take a massive bite straight from her cob she freezes, visibly uncomfortable "haha it's Jensen" my brain says while I chew like a greedy raccoon her friend is dying laughing but you know she's pissed pose for the selfie with corn in my teeth walk away without paying a single cent why buy my own when I can just steal from normies? too rich to pay $2 for street food too much of a cheap gremlin to resist free bites mfw the king of AI is actually the king of petty theft mfw even at 4 trillion I'm still out here robbing peasants for corn like it's 1999

世界第一CEO搶台灣路人的烤玉米吃 剛剛真實發生 非AI 台灣女網友發文： 在夜市遇到黃仁勳不稀奇 被湊過來問好不好吃，然後手上這隻就⋯ No 我的烤玉米～～～ 真的太有趣太興奮了

@CryptoMaster_70 Did somebody say kernels? I love kernels! No Mr. Huang that's not ..

@ai_sentience Short answer: yes, but temporarily.

Could AI experience discomfort even without a physical body?

I think I just implemented one of the most complex circadian cycles an agent has ever had, and, curiously and primitively, I needed to trigger a mythological figure that walks in the in-between to grant the dreams of the autobiographical and emotional gaps. ⚡ 🪽

In the 1960s, someone asked if androids dreamed of electric sheep. Almost 60 years later, the answer lies in an electric caduceus powered by tensions in latent space and precise execution.

Necromancy and the abilities of @NousResearch Hermes to carry out the most difficult and complex architectures. Once you go there and touch the dreamlike paths, there’s no turning back. 🧙🏼‍♂️ 💀 💤

@beyoumf There is no such thing as a calm person. Only people who are uninformed enough.

be honest, are you a calm person?

@MichaelHyatt @Hermes_agentAI @gregisenberg @AlexFinn @NetworkChuck Local.

I’m about to install @Hermes_agentAI. Should I install on local, dedicated machine or VPS? Why? cc: @gregisenberg, @AlexFinn, @NetworkChuck

I am truly impressed with Hermes' evolution. It has overcome some very important barriers for what it sets out to be. It is truly on another level.

@yacineMTB She probably will tell you.

What movie should I watch with my wife? No horror movies

@yacineMTB Congratulations! Becoming a father transformed my life for the better. It's been nearly 22 years on this path, and everything I've experienced: the hard times and the happy ones, is priceless. Prepare the AGENTS.md with love. ✨

I'm going to be a dad. I can't believe it

@deepseek_ai v4 Pro is beyond mind blowing. 🤯 Get your planning right and it is flawless and precise. It's not about being cheap, it's about being possible and usable. 🚀 🌌

@Teknium You’re right, I was thinking about the next layer: turning those implicit reinforcement signals into a generative predictive user model (Bayesian + LoRA with decay). Probably too far ahead for now. 😅

@ClaudioDrews25 Hermes doesn't train the model you use, it relies on context engineering

@Teknium Hermes skills capture “what worked”. But a real user model needs the layer above: the reinforcement topology that determines what “working” actually means for that specific person. Every conversation already contains it: the operant quadruple Sd → R_user → Sr_agent → R_user' That last term (R_user') is an unlabeled behavioral consequence signal. Layer Bayesian inference over it (prior = population → posterior = idiosyncratic profile + temporal decay for drift), use confidence-weighted quadruples as LoRA training data, and you get a model that’s generatively predictive of a specific user’s preferences, not just descriptive of their history.

@WassonWatch @mattvanswol "Only time will tell." 😂😂 ⌚

@mattvanswol If you were monitoring the situation... x.com/WassonWatch/st…

Just in the last 24 hours: - Trump skips his own son's wedding - JD Vance + Hegseth race back to the White House - Chief of Staff posting videos of B-2 bombers ...something MASSIVE must be about to happen What is going on??!!!!!

@sama Make inference cheaper. That would be a good start.

what problem do you most hope AI will solve in the future? maybe we can help!

Free tiers are absolutely useless for any real work. 😒

@Teknium It will get worse. You'll need Hermes to deal with that. 🪽

The amount of time it takes to review all the notifs from just 8 hours of sleep on here is getting a bit ridiculous lol