MindVenture

Rage, rage against the dying of the light.

@Card_Cap Being too early can be as bad as being too late though

If you’re on time, you’re late. If you’re early, you’re on time.

@GraemeVIP @claudeai Keep in mind, what this means is that OpenAI is just losing a lot more of their investors money on you. At some point, all of this will stop even for Codex. OpenAI is burning money by the boatload.

Allow me to fix your post, "You can now run multiple Claude sessions side by side for around 5 minutes before we lock you out and demand another 200USD off you". I can burn through the 5h limit in about 8 minutes. That's absurd in anyone's book. I can absolutely hammer Codex for hours (same tasks) and it doesn't run out. Both on the same plan. How on Earth is anyone getting work done with Claude? GPT5.4 is a more advanced model than Opus 4.6 right now anyway - especially since they nerfed it for the Mythos/4.7 release.

We've redesigned Claude Code on desktop. You can now run multiple Claude sessions side by side from one window, with a new sidebar to manage them all.

@parmita If what you need is an answer which uses the latest information, use Perplexity or Gemini. Never Claude.

I asked Claude to choose the best flight route “given the war.” It searched for the Russia Ukraine war…It still lives in a time when “war” meant Russia and Ukraine…I didn’t even correct it. Let the model dream.

@chatgpt21 All features are not equal tho

I went through every official changelog and product announcement post for the major AI labs for the past 2 weeks (April 1–15) The rules I set were: user-facing only, chatbot and code editor surfaces only, no API pricing tiers, no model deprecations, no backend infra. Distinct features only, not release events. Anthropic: 13 OpenAI: 5 Google: 4 xAI: 1 The shipping velocity gap between Anthropic and everyone else is real and it's widening, and mythos is obviously accelerating this.

@crftrwrnch @BrianRoemmele FBI or CIA may be the very ones doing these "disappearances" so ... why investigate themselves you know?

@BrianRoemmele Does the FBI even bother to investigate stuff like this anymore or all of them busy with blackmail and extortion assignments?

Oh now at number 10.

@kimmonismus I do like the idea of users picking themselves, but I have a feeling that the model should be automatically chosen based on the input for the best result.

Antropic is rolling out switching models mid-chat. About time.

@cjwestland @bentlegen Everything big enough. If you have a nice small niche and are bootstrapped you're probably good for a few milly. Anything in the hundreds of millions or billions per year will get gobbled up.

@bentlegen I literally had a slide that I showed my team with Amazon batteries and my purchase history, raising a flag about "Anthropic never gonna build this; we're too niche." It's a huge, huge, important lesson from the past; they are going to go after everything exactly like you said.

I don't think people understand that Anthropic is like Amazon now They can just look at the numbers on their platform and decide which products are worth replicating

@nanalyzetweets Yeah, that's true. They'll probably IPO at an insane valuation that doesn't make any sense.

@Mindventure_ Folks all find these late stage startups to be so appealing but the reality is that by the time we're able to invest, much of those exponential returns are already realized and we're just someone else's exit liquidity for said returns :)

This is a crazy metric. Revolut now has more users than $SOFI $HOOD Dave and Chime combined.

@MisterMorrill @doodlestein @sama @sarahdingwang The world is run by psychopaths and we should really do something about it. That guy had balls and went for it. Kudos.

@doodlestein @sama @sarahdingwang Reading reddit posts about United Health exec murderer on Reddit is disturbing. So many people think he is a hero.

I wrote this early this morning and I wasn't sure if I would actually publish it, but here it is: blog.samaltman.com/2279512

@Beyond_Scarcity @sama Disagree, Sam is a despicable human being and has already hurt thousands if not more people.

What happened here is abhorrent. Violence, intimidation, and attacks on someone’s home are never acceptable, and I’m very glad no one was injured. People have every right to criticize, disagree, protest, argue, and speak openly about AI, the companies building it, and the people leading them. Free speech matters, especially in a moment like this, when the stakes feel so high and the technology is moving so fast. Strong opinions, public scrutiny, and open disagreement are not the problem. They are part of a free society. There is also nothing unreasonable about wanting safety, accountability, and serious debate around AI. Those concerns are legitimate. People are right to ask hard questions about power, control, risk, and who gets to shape the future. But none of that extends to violence. There is a fundamental difference between expressing an opinion and throwing a Molotov cocktail at someone’s house. Defending free speech means defending the right to speak, criticize, and dissent. We should be able to have fierce disagreements about AI, its trajectory, and its consequences without treating each other as enemies to be attacked. In fact, if this technology is as consequential as many believe it is, then open speech, honest criticism, and vigorous public debate are even more important, not less.

@xdNiBoR Technically, we are already out there in the stars even while on Earth :D

This will be real one day

@BradoCapital @PrincipatusCap What would you recommend for someone who needs to pull data for 10 to 20 stocks for his personal portfolio tracker?

@PrincipatusCap we do have individual pricing for this stuff, but you do need to fill out the form as an individual

Fiscal is now an official Claude Connector. 🥳 Get our structured data content into Claude with one click of a button if you have the API & MCP plan. Fiscal is now officially partnered with most frontier labs to deliver accurate & timely financials to investors.

@PDoherty16 @mikekatz29 Yeah, how does this work? If the lawyer uses cloud solutions and AI, it still goes to the cloud and Anthropic or Open AI servers.

@mikekatz29 as lawyers all of our data is stored on third party servers - Microsoft, Dropbox, etc. No one takes the position that the A/C privilege or work product does not attach to those communications. At a minimum the scope of this ruling will have to be limited over time.

The epidemic of vibe lawyering is going to cause some problems. A friend recently typed the details of an employment dispute, including things he did in violation of his non-compete, into ChatGPT and asked it to draft a separation agreement. He did not talk to a lawyer first. He created a discoverable record of his own liability that lives on a third party's servers. He is very much not the only one. I wrote about why this is a real problem, and what a federal court just confirmed about it.

@realroseceline Cna you elaborate more?

This is why banks are NOT INVESTABLE, regardless how good you think the biz is, how fast it’s growing, how well you think you understand it, or even how cheaply it’s trading for. Avoid complex business models and look for simple ones.🌹

@Unfiated @KobeissiLetter Ignore Trump and think about Netanyahu. He is the one who actually makes the decisions, not Trump. As soon as you realize this, everything falls into place and makes sense. BB doesn't want a deal. He wants a destroyed Iran.

@KobeissiLetter So he destroys the Middle East and asks those destroyed countries to pay for it. This guy is a clown 🤡

BREAKING: President Trump is interested in calling on Arab countries to "pay for the cost of the Iran War," per White House officials. President Trump will soon have more to say on the topic, according to the White House.

@LamboParsons @IterIntellectus Of course we have it. It's just not as profitable :)

ok so the rosie story was even more insane than it looked > be the australian tech guy who made a cancer vaccine for his dog > first try: genetic algorithms to design a new drug from scratch > works in simulation but would take years to test > second try: screen 1 million existing compounds against the mutation > two weeks of computation. find a perfect match > it's patented > patent holder says no to compassionate use > what_did_you_expect.jpg > spend two weeks just being with the dog > 2am idea: what if i just make a vaccine > chatgpt for pipeline, gemini for construct, grok for validation > 300 gigabytes of raw sequencing data to half a page of vaccine construct > university ethics approval would take until mid-2026 > dog doesn't have that long > panik > canine cancer expert connects him to a lab in queensland with existing approval > drive 14 hours to get there > inject > three weeks later the tumors swell. immune system swarming > six weeks later shrinking > two months later legs returning to normal > one mass doesn't respond > sequence it again > different cancer. the vaccine worked. the body grew a new tumor he's now building a company so every dog owner can do this he had the technology the whole time. he spent 18 months fighting for permission to use it

@FWNewsOfficial @KobeissiLetter NATO is a defence alliance, not a war alliance.

Iran recently granted Spain unrestricted safe passage across the Strait of Hormuz, but Spain continues to oppose the US using its bases against Iran. You can't stab your NATO allies in the back during a war and expect protection when you need it, as Sánchez refers to it as avoiding "Russian roulette."

BREAKING: Iran has granted Spain safe passage through the Strait of Hormuz as Spain continues to prohibit the US from using its military bases for operations against Iran.

@netcapgirl Really hope so. But what will humans do?

claude cowork is making me think maybe we’ll look back and it’ll be obvious that humans were never meant to spend their lives working behind a screen. we’ll see it as inevitable that computers do everything for us on computers and the future of work is cooler than we can imagine

@KayvonJafar @buccocapital AI slop detected.

@buccocapital yeah this tracks. most models “learn your voice” by averaging you into generic ai prose. claude actually picks up your constraints, cadence, and what you don’t say. the real flex isn’t the rewrite. it’s the instruction set that makes the rewrite repeatable.

I uploaded like a year of my writing to ChatGPT and Claude and asked them both to generate instructions to write like me ChatGPT output is legitimate trash. Garbage slop that is clearly written by AI Claude? It literally writes just like me. Pretty unbelievable