cyclistal

@SterlingCooley Still burning up tokens on the back end. THAT is what gets expensive. Sure, one can use open models or older frontier versions, but the FOMO of using the latest and smartest is hard to say no to.

Why don't people just make their own servers ? digital ocean is pretty damn easy, and you can literally install Claude Code on it c'mon guys

@chadwahl He was feeling fancy so he wore a tie. We’ve all been there.

Fetterman looks like an FDE walking into an executive update after building all night and final merge 30 seconds before the demo

@a16z @qasar Where are his shoes?

Qasar Younis on how Applied Intuition uses books to mold company culture: "We also read books as a company, as a leadership team, and then we invite the whole company to do it." "Think about your company as this bespoke piece of clothing that you get made, and then you grow, and then before you know it, you have to constantly re-tailor your company for the scale and type of company it is." @qasar @AppliedInt

@RibierCedric @DuskNixx Oh, so you’re going to be like that, eh? Well then, your mother was a hamster and your father smelt of elderberries!

@DuskNixx @cyclistal Ouais tu fais rien du tout et tu demandes l’autorisation à ton SysAdmin.

Si vous possédez un Mac, rendez-vous service et n’exécutez pas la commande qu’un random X/Twiter vous donne sans contexte. Voilà.

@SMB_Attorney “Make it airtight” is the new “Make no mistakes” 😂

> Your boss asks you to lead a project > You need a legal agreement > You open your LLM of choice > “Draft me an agreement.” > “Make it airtight.” > “Add a custom provision for this weird edge case.” > LLM delivers a 25-page, hyper-protective agreement > You don’t fully understand all of it, but it sounds right > You send it > Counterparty opens their LLM > “Find every issue.” > “Rewrite this in our favor.” > “Also add protections so we don’t get burned.” > You get back a 35-page redline > Half the comments contradict yours > Some provisions now interact in ways you don’t understand > Your boss asks: “Are we covered here?” > You pause Because now the real question isn’t: “Can an LLM draft an agreement?” It’s: “Do I actually understand the risk I’m signing up for?” And: “If this goes sideways… who owns that decision?” What’s your next move? Do you get it now??

Pete is the GOAT. Helpful to all and ships features and fixes on the regular. Hermes has a following (and bots) posting on just about all the OpenClaw related messages on X. No shade to them, but I’m backing the Claw due to it’s virtuous cycle of great support and massive adoption.

by far the most painful part of using hermes (and openclaw) is secrets management. who is solving this?

@iamlukethedev @Scobleizer Is the memory good enough to replace QMD/LCM?

OpenClaw v2026.4.10 just dropped 🦞 This one makes your agents smarter in conversations and way better at remembering things. Plus upgrades across video, voice, and reliability. Here’s what actually matters ↓ 1. Active Memory plugin This is a big one. Your agent now remembers for you. It automatically pulls: • Preferences • Context • Past conversations No more “remember this” every time. Conversations just flow naturally. You can inspect it with /verbose and tune how much memory it uses. 2. Codex provider Codex now has its own native integration. • Managed auth • Native threads • Model discovery Clean separation from OpenAI models. Better for dev-heavy workflows. 3. Local speech on macOS Talk Mode now supports local speech. Runs fully on-device. • Faster • Private • Handles interruptions Huge for voice-first setups. 4. Video generation upgrade Seedance 2.0 now supported. Better control over: • Duration • Resolution • Audio • Output consistency Agents can create better video now. 5. Microsoft Teams upgrades More control inside chats: • Pin and unpin • Read • React • List reactions 6. Small upgrades that matter • Live streaming messages in Matrix • New exec policy CLI commands • Better support for self-hosted models • Cleaner dreaming UI • Improved behavior for GPT-5 agents Plus a lot of security and stability improvements. Safer execution. Better startup. Fewer edge-case issues. No major breaking changes. Smooth update. How to update openclaw update openclaw –version npm install -g openclaw@latest Docker: pull latest image and restart This one is important. Agents are starting to feel like they actually remember and think.

@Noahpinion This attack occurred well before Mythos was created. But your fundamental point is still strong.

Just today, someone asked me: If our government knew about Mythos ahead of time, why didn't it team up with Anthropic to hack China before announcing the model's capabilities to the public? 😈

@wallstengine Third verse, same as the first!

Axios: OpenAI is planning a staggered rollout for a new model with advanced cybersecurity capabilities, limiting access to a small group of companies over fears the tool could be misused. The move would mirror Anthropic’s restricted release of Mythos.

@AndrewBenson @8teAPi Sure, but Sam and Greg will be kicked out. That’s what this is all about.

this is getting so complicated. OpenAI could lose the lawsuit but still the current foundation ends up increasing its stake and Elon still doesn’t get control

@twistartups “Find the bugs and patch them” sounds great, but one never finds all the bugs in one shot and systems keep changing so bugs keep popping up. I can’t ask an LLM to “fix all the bugs in my software. Make no mistakes.”

Sam Altman has alienated so much of his top talent at this point… He’s driven a ton of people out of OpenAI, including Dario Armodei, whose company, Anthropic, is now beating them in many ways. The AI race isn’t just benchmarks — it’s leadership. Dario is clearest example, but it seems like more and more top employees at OpenAI are stepping away all the time…

@allie__voss Next thing you know they will build a wall to keep everyone in!

States worth living in don't have to threaten people into staying

@AndrewCurran_ They should have let Mythos represent them. Live and learn!

A federal appeals court in Washington DC has denied Anthropic's request for a stay in its lawsuit against the Department of Defense.

Your dog could soon get a daily pill that slows aging. 1,300 dogs are already testing it across 70 vet clinics, and it just passed its second FDA safety review in January. The drug is LOY-002. A San Francisco startup called Loyal makes it. Celine Halioua started the company in 2019 when she was 24. They have raised $250 million so far, including $100 million in February from Baillie Gifford, the same firm that bet early on Tesla and Amazon. The simplest way to explain what it does: if you put a dog on a strict diet where it eats way less food, that dog tends to live about two years longer. Cancer and arthritis both show up later. The body just ages more slowly. LOY-002 gives the dog those exact benefits without actually eating less. The dog keeps its appetite and its weight; it just ages more slowly. The trial is called STAY. Half the dogs get a beef-flavored pill every day, and the other half get a fake version that looks and tastes the same. The FDA reviewed safety data from over 400 dogs. Zero serious side effects, even at five times the normal dose. Loyal has now cleared two of three hurdles for conditional approval, which basically means they could start selling the drug while the full trial keeps going. The last piece, manufacturing, goes to the FDA in 2027. One thing the tweet gets a bit wrong: the pill is for senior dogs aged 10 and older, at least 14 pounds. And Loyal is being careful about the "years" claim. Right now, they are saying at least one extra year of healthy life. They want to prove more, but the trial needs to finish first. If LOY-002 gets approved, it would be the first drug the FDA has ever cleared to slow aging in any species, human or animal. Loyal also has two more drugs coming for big breeds like Great Danes and Rottweilers. Those dogs die younger partly because of a hormone called IGF-1, which makes them grow fast but also wears their bodies out quicker. And a separate project backed by a $7 million grant from the NIH (the agency that funds medical research) is testing a human organ transplant drug called rapamycin in 580 dogs to see if it slows aging, too. The global pet care market hit $261 billion last year. A once-a-day pill that gives your dog even one extra healthy year could easily become one of the best-selling veterinary drugs on the planet.

@lukOlejnik @hackerfantastic Wait… I just realized that APT-Actors also offer HPaaS once they have a foothold. It prevents others from getting in and messing up their game.

@lukOlejnik @hackerfantastic Hack and patch as a service. HPaaS But only the government can offer HPaaS

“We hacked your router to protect you” has become a repeatable law enforcement tool, the courts keep saying yes, and this is fine.​​​​​​​​​​​​​​​​ The FBI broke into home routers using the same bug the Russians used, armed with a court order instead of a Kremlin directive. The FBI patched the routers. Reset the DNS, blocked re-entry, pulled forensic logs. The third time they’ve done it (Exchange servers in 2021, APT28 botnet in 2022).

when banthropic detects the openclaw gateway on your system

@KaiXCreator Preach! The API costs are killing me.

The more I use OpenClaw, the more I think it's not just hype. The only thing limiting me now is the API costs.

@KentonVarda I don’t know what all the fuss is about. Just don’t let the bad guys use it.

Honestly "AI that can find every vulnerability" sounds way better for the good guys than the bad guys. Not sure why everyone is losing their minds here.