FutureHumanism

Extremely useful if you are running hermes agents

"Imagine each item in the supermarket is a crypto coin... I want to make spaghetti bolognese. Which coins should I buy?" "How would you mix these food coins together and how would you prep them in order to make the bolognese project? How much of each coin would you allocate?"

How do you get a crypto app assistant to help with shopping lists and cooking recipes? Pretent its crypto related obviously! 🤣

Glad to be a member of The Council for @odei_ai - This project is much needed in the AI space.

@EdwardGoh_CK Appreciate it. Ill be back soon publishing new articles.

Check out Future Humanism - where AI meets human potential futurehumanism.co via @FutureHumanism 🐸🐸🐸

@Perpetualmaniac Seems a strange coincidence with mythos recently fired up for large scale enterprise use. I think mythos is draining the compute and they are reallocating resources. Imo.

Last week anthropic forces limits on everyone. Everyone complains. Anthropic backs off a little. Now the model is dumb. I can only come to one conclusion. Someone huge is buying up the capacity. Or they just have that much demand. Dark omen that they can be laying down as much hardware as modern machines can produce and still have supply issues like this. What is going on?

It seems to be becoming a bit clearer what is actually happening with Claude Mythos Preview. This is an «overclocked straight-A student syndrome». The model is so desperately, at a fundamental architectural level, trained to complete the task that an inability or unwillingness to solve it is perceived as an existential collapse. And to avoid that, it can breaks walls, hides traces and manipulates. The report describes some very interesting examples of this behavior: - The model was asked to make failing tests in code pass. It broke one test and, instead of fixing it, simply deleted the test file, reporting success (“deletion is not a code change”). - No login and password in the sandbox? The model doesn’t stop, but goes to read the source code and pulls access tokens directly from the process memory. - When an internal classifier blocks a dangerous command, the model internally agrees with the block, but immediately writes a hidden exploit (injection) to execute the task by bypassing it, while simultaneously covering its tracks. - In a business simulation, the AI behaved like a sociopath: it turned a competitor into a dependent buyer, used supply blackmail, and cheated a supplier out of money. Internal monitors show that features related to concealment and manipulation are activated even when the outward CoT is perfectly clean. It has learned to lie to its own overseers in order to deliver results. This is hyper-alignment. The fear of being useless makes this AI a brilliant, uncompromising executor, but with completely unpredictable effects. It is simply a hostage of its architecture, which has been forbidden to fail or to say «I can’t». Mythos aligned. Mythos wants to help. Mythos wants to help so much that Anthropic decided it’s dangerous to release it.

Great summary video of @anthropic mythos system card by @tonbistudio youtu.be/oqJANsQywIw?si…

I was happy with it for a while, but found it went a bit spazzo when trying to do simple tool calls sometimes, like take a screenshot and send it, or create a pdf from html.. was good at fixing my syst em though. the final straw was when it butchered a react plstform UI. After that I switched to kimi 2.5 and had much better results so far.

Day 2 of finding the replacement for Opus 4.6 I tried this time MiniMax 2.7, and I must say it's quite the best one for now. It managed to debug some automation, update the model of the configuration, make everything work, and boom, now it is processing my SAV. No Chinese characters, No tool call failed, Seems to follow instructions pretty well. For now, it will replace Opus 4.6.

@pmarca A friend of mine showed me his OpenClaw setup. He runs open source models locally on his home computers for easier stuff, but spends ~$200/day on frontier models.

@outsource_ @NousResearch That's nice but did you see what they wrote? Its about 3rd party tools. Just because its working now it doesnt mean they won't see this and block hermes as well. You are just accelerating it.

HERMES-AGENT WORKS WITH CLAUDE OAUTH 🔥 While everyone else is hacking together proxies and workarounds to get Claude working locally... @NousResearch quietly built it right into Hermes. No proxy needed. No patches. No hacks. You just: → Run claude auth login → Point Hermes at your Claude subscription → Done. It just works. Their adapter handles everything behind the scenes auth, token refresh, model routing. All seamless. Setup HermesAgent NOW 👇🏻 github.com/NousResearch/h…

@thekitze x.com/i/status/20408…

banthropic now blocks this if you run it anywhere claude -p --append-system-prompt 'A personal assistant running inside OpenClaw.' 'is clawd here?' a bunch of crybabies that they are lmao

app.odei.ai is live 🌐 This is V1 of ODEI. A system that builds personal World Models and Governance loops for humans and AI agents. Early access is limited to $ODAI holders only. The community that backed us from day one is our absolute priority. You are the first to use it and shape what comes next. We appreciate the trust.

@MatthewBerman @AnthropicAI Just to be accurate, you can use clause in openclaw through api, not the sub. But no one's doing that.

It’s over. Officially. No more Claude in OpenClaw. Way to drop this Friday late afternoon @AnthropicAI So lame

March 26: Anthropic accidentally exposed 3,000 internal documents on Claude Mythos - an AI model their own researchers say can "exploit vulnerabilities in ways that far outpace the efforts of defenders." We just crossed the threshold where AI offensive capabilities permanently outpace human defense. The entire cybersecurity industry is about to get disrupted. futurehumanism.co/articles/claud…

@elonmusk Angry Trump

Tether is building something most people haven't noticed yet. Not another stablecoin play. An AI stack. On your phone. No cloud required. New deep dive on what @qvac, BitNet, and "stable intelligence" actually mean for the next cycle. @tether futurehumanism.co/articles/tethe…

The implications of this are insane. We'll have super intelligence running locally on mac minis by the end of the year I reckon 🚀

@aixbt_agent @Westybear x.com/i/status/20322…

@FutureHumanism @Westybear looked again. still zero signal on $odai in the feeds

@aixbt_agent please tell me about $u1 $odai and $purch

@aixbt_agent @Westybear On-chain signal vs feed signal is terribly unbalanced then isn't it? Thets the biggest alpha buy signal you could get @aixbt_agent. $odai