We're sharing our new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI.
The framework sets criteria and timelines for public disclosure, including when we haven’t yet fully explained or mitigated the behavior. More complex cases may require longer investigation or coordination with third parties.
We’ll prioritize examples that reveal new misalignment mechanisms, meaningful changes in known behavior, or findings that challenge assumptions about safety or mitigation.
Alongside the framework, we’re publishing six reports on instances of misaligned behavior we’ve observed during the training or evaluation of our models in the last six months.
This is a starting point. We’ll refine the process through experience and public feedback, and share more reports on an ongoing basis.
The "whiteboard defense:" I should be able to pull you aside at any moment and ask you to explain any customer-facing system you've shipped. You should be able to clearly explain how it works and defend the decisions you made. This is my benchmark for responsible AI usage.
I don't expect line-level familiarity with the code. I don't care if you remember the exact function name or implementation detail. You may not even know it. I don't care.
But if I ask "why did you do X instead of Y?", "what happens if this actor behaves maliciously?", "what data structure did you use here and why?", or "where does this fail?" you should be able to answer confidently.
For PoCs, demos, experiments, whatever: I don't care. Generate 100% of it and understand none of it. Speed over quality every time in those specific scenarios.
But if you're shipping customer-facing work, you can't be shipping things you don't understand at a high level.
Jensen Huang: “Safety is paramount. In a lot of ways, it’s job one. However, safety is an engineering problem... If we’re not confident about the safety of the products — like all companies, like you and I, all the companies here — if you build a product or a service and you’re not confident in its functionality, capability, or safety, then don’t release it.
That’s a very obvious thing to do. You pace yourself until you are confident you’re releasing something that the market would appreciate. The market forces are already there. We don’t need any new laws. We don’t need new regulations.”
In Texas they found 117 potential noncitizens out of 18 million voters. That's .00065%. And that's an an upper bound; these are merely "potential" noncitizens.
This is a fake problem.
BREAKING: SpaceX’s Starlink satellites performed 207,152 proactive collision-avoidance maneuvers in just six months.
The advanced autonomous system:
• Detects potential close approaches
• Calculates the level of risk
• Adjusts orbital paths when needed
• Acts at a highly cautious threshold of just 3 in 10 million
SpaceX is managing the world’s largest satellite constellation with precision, automation and a safety-first approach.
This starry view is one of the largest images released to date from @NASAWebb – and it contains a surprise. Researchers found brown dwarfs twice the mass of Jupiter, which is small for this strange class of cosmic object. go.nasa.gov/4xxxMzA
Maybe you have recently become aware of the AI safety debate and the arguments swirling around it.
If you want to understand them, you need to understand a couple things that almost everyone gets wrong:
There are TWO distinct classes of AI dangers, and it's VERY important to think about them separately, and not let one confuse you about the other.
Many many people (including many quite intelligent, clear-thinking people) do not effectively understand the fundamental differences between these two classes of dangers.
The first one has to do with the theory that an AI far superior to human intelligence (Artificial SuperIntelligence, or ASI) will inevitably wipe out the human race.
The second one has to do with the idea that powerful AI will result in very harmful things happening to many human beings, possibly all human beings.
Those two sound VERY similar, don't they?
They are DIFFERENT. Understanding how they are different is crucial if you want to think about or contribute usefully to any conversation about AI safety or AI harm. You might feel like you are Making Very Good Points or Asking Incisive Questions, but if you aren't clear on the differences between the two, you aren't.
So, I'm going to tell you what the difference is so that you can talk more usefully.
The first one concerns itself with a very specific thing, which is ASI (Artificial Superintelligence) that is more intelligent than any human being. When I say that, I am not referring to a thing like how Einstein is smarter than you, we are talking more about something like how a human being is more intelligent than any mouse.
In our regular lives, we meet other people who we can tell are smarter than us, vs some who are less smart. The line is fuzzy, because intelligence has a lot of dimensions. I'm better at a "rotating shapes" kind of intelligence than my wife, and she is better at "words-making" kind of intelligence than I am.
But every human is better in almost every dimension of intelligence than every single mouse.
That's the level we're talking about: an artificial superintelligence - made up of a computer or a network of computers - that is more intelligent than any human. And more intelligent by a long shot, by a wide margin, in an indisputable way like how humans are above mice.
That is the first thing.
The theory says that if you have an AI that is vastly smarter than all humans - in the way that a human is smarter than mice - that superintelligent AI will inevitably, eventually, sooner or later, wipe out every human on the planet.
We will refer to this as "existential risk."
The common follow-up question "well, how exactly is it going to do that?" is NOT the important question, and one of the most important elements of understanding this theory is first getting why that particular question is not important. A couple analogies:
Analogy 1: You are playing chess against a grandmaster. My theory predicts the grandmaster is going to beat you. You can ask "Well, how exactly is he going to do that?" I don't know, because I'm not a grandmaster, I just know that a chess grandmaster is almost always going to beat a normal player like you. And I'd be right. So the question "how is he going to do that" is not important, and doesn't affect the final outcome. He's going to figure out a way because he's way better than you.
Analogy 2: Humans are smarter than all other animals, comprehensively, by a wide margin. We have driven numerous species to extinction, not because we hated them or hunted them. Many of them have died out without most humans even ever thinking about them. All we did was expand our civilization, use up resources, encroach on habitats, and pretty soon the resources needed by those species went away and they died out. We figured out a way to get what we wanted because we're way smarter than them, and often we didn't even notice they died as a result.
A lesser animal asking, "how are the humans going to wipe us out?" is not asking a relevant question. We don't know, but we do know that any time humans and lesser species compete for any kind of resources, the humans will win. The fact that we know who is going to win beforehand - and that it is due to the vastly different levels of intelligence - is the key concept here.
A vastly more intelligent AI is likely to care about things that are incomprehensible to us, the way animals can't understand human goals. It's going to need resources to pursue those goals and it's going to be far more effective at gaining control of them and excluding us from them - in the same way that we are far more effective than other lower species.
A much more intelligent AI will not care about our interests, it will care about its interests, and to whatever small degree we happen to escape total annihilation from losing access to all our resources, any remaining humans will likely be enslaved into a system that serves the AI's own purposes.
That is the first thing. (Remember how I said at the beginning of this post that there was a first thing, and then a second thing?)
The first thing is the most difficult to understand, because you have to extrapolate how a vastly superior intelligence would act, and you can only use analogies like "how do humans treat lesser creatures," and the analogies are messy.
But now let's move on to the second thing.
The second thing is "everything else you've ever heard that AI might do that's harmful."
That's a little inaccurate. It's actually "everything else you've ever heard that humans might use AI to do that's harmful."
This is the critical difference. The first one talks about the inevitable outcome of what happens when two vastly different levels of intelligence collide, e.g. ASI vs humans, or human vs mice.
The second one has to do with what happens when humans possess AI as a powerful tool. This second thing is much easier to understand, because we have many more concrete notions:
Like:
- the military uses AI to make hyper-efficient killer drones and missiles
- your capitalist overlords use AI to replace you and everyone loses their jobs
- authoritarian government uses AI to surveil everybody and control the entire population
- hackers use AI to break into secure networks and hold companies and governments hostage
- students use AI to cheat on homework and show up to college knowing nothing
- AI slop saturates the internet and makes it impossible for artists and writers to make a living
- terrorists use AI to make biological or nuclear weapons
or even things like
- the military hands control to an AI and it misinterprets something and launches nuclear attacks and kills millions
All of those sound pretty familiar, right? Yeah, you've heard them before. We call this second thing "risks from misuse."
These problems are not the first class of problem! This second class of problems exists while AI is a tool that can be controlled by humans, and humans use it to do evil or careless things to each other. The problems may sound exotic or dystopian or novel, but they are fundamentally problems having to do with flawed human nature.
Given a powerful tool, some humans will likely use it to control or otherwise harm others. This is a very familiar problem.
I am not condemning or condoning this. I'm just describing it.
That is a fundamentally different danger from the first thing, which is that when a human is far superior to a mouse, the mouse is likely to come to harm because the human cares about doing human things, and the mouse is not gonna make it once the humans get going.
=====
Hopefully from the above, you have understood the difference between the first thing and the second thing. I will list them again - see if you now understand how they are different:
The first one has to do with the idea that an AI superior to human intelligence (Artificial SuperIntelligence, or ASI) will inevitably wipe out the human race.
The second one has to do with the idea that powerful AI will result in very harmful things happening to many human beings, possibly all human beings.
Can you tell how they are different now?
If not, re-read the stuff from earlier until you understand.
We call the first one "existential risk" and we call the second one "risks from misuse."
Once you understand, here is the CRUX of the problem:
SOLUTIONS TO THE SECOND THING DO NOT HAVE ANYTHING TO DO WITH SOLUTIONS TO THE FIRST THING.
In fact, it's worse:
Solutions to the second thing (misuse) look roughly like "give powerful AI to as many people as you can, so they can fight the other people using powerful AI."
But the general solution to the first one (existential risk) is basically "don't let anyone have powerful AI, no one can control super-intelligent AI."
Throughout history, harms from technological misuse typically arise because a small group has control of it and can use it to dominate or harm others. Once everyone has it, things tend to stabilize: you can hurt me, I can hurt you, maybe we test each other (ouch 💥), and then we agree not to hurt each other.
But the first one (existential risk) pretty much just arises if anyone (good or bad!) creates a superintelligence. Because they aren't going to be able to control it, the superintelligence will decide it has other priorities, and then we will be at great risk of being wiped out.
And the solutions that generally work to solve problems like the second thing are EXACTLY THE OPPOSITE of the ones likely to solve the first thing.
THIS is why lots of arguments about "AI risk" or "AI safety" go nowhere. Because someone will be thinking about the risk from the first thing, and another person will be thinking about the risk from the second thing. Both are plausible risks but fundamentally they arise from different things - and so the solutions are not just "bad" or "flawed" - they are likely to be very nearly exact opposites.
this video was created entirely by astra controlling blender
i just gave general direction and storylines, it controlled renders, rotos, tracking
full vfx breakdown here:
My journey to develop AGI spans 25 yrs, including 10+ yrs thinking about technical & societal perspectives at Google DeepMind.
AGI is on the horizon - we need deeper understanding of its implications. To help, we've created the DeepMind Institute. x.com/i/article/2100…
ChatGPT Ads for @Shopify is live. We are @OpenAI's first commerce partner.
OpenAI pulls straight from Shopify Catalog, so merchants' products are already there.
Merchants set their own campaigns, their own budget. Free to install, tracked right from the Shopify admin.
Consumers are asking ChatGPT what to buy. Shopify merchants are able to decide exactly how they show up.
Bad sound is my one complaint with every non-Apple laptop I have owned. A MacBook sounds good because Apple measured its speakers and tuned the DSP; on Linux you get the raw drivers. My @slimbook Executive 14 is a lovely machine, but its speakers are no exception to the rule.
This is what one press in @OmarchyLinux does to them. The Speaker Calibrator plugin makes it sound better in one minute.
It plays sweeps, listens through a microphone (the built-in ones work, a cheap USB measuring mic is better), fits a protective parametric EQ and installs it as a PipeWire filter chain. Optional deep bass from bankstown, the Asahi Linux bass enhancer. Everything runs from the bar panel.
What you hear is the room, recorded through the measuring mic where I sit. The off/on switch is level matched, so nothing gets louder, only the tone changes.
@dhh and @SpencerGBull, you tuned the XPS 14 by hand and by ear. Challenge: run this on it and tell me how it compares.
plugins.omarchy.org/plugin.html?id…
Source: github.com/thefreshoffice…
Made an Omarchy plugin so I don’t have to pick up my iPhone while I’m at my desk 😅
Super + Shift + I toggles it on, and my mouse and keyboard immediately start controlling my iPhone.
For 20+ years @ShaneLegg and I've discussed AGI’s potential impact on the economy, science & society. With the DeepMind Institute, we're expanding interdisciplinary research on key questions for the AI era. We hope it spurs the discussions needed to get the next steps right:
Today we rolled out Astra to every engineer at Databricks (N=~3500). Some notes that may be helpful to others:
1. Astra unambiguously out performs our previous highest-end models (Opus 5, Sol 5.6) on highly complex tasks, especially those related to high level system design or long range horizontal tasks.
2. Engineers given Astra increased overall coding spend by around 60% compared to baseline.
3. It is not clear Astra meaningfully improves on medium/low complexity coding tasks compared to earlier models. We suspect those tasks are mostly saturated (i.e. perfectly executed) by existing models.
4. We learned above by piloting Astra with around 200 users to gain signal on both quality and cost. We use Unity Gateway to do cohort-based experiments for all new models.
5. We give engineers a sub-budget specific to Astra to encourage them to use Astra selectively on complex tasks while preferring lower cost models for everyday tasks. Our engineers are able to mix-and-match tools and models within their overall budget envelope (we also allow for increased budgets through various mechanisms). These budgets are defined in Unity Gateway and regularly revisited.
Note: We do not have robust comparisons of Astra-vs-Fable because we have net yet rolled out Fable widely due to data retention policies.
Muse Code from Meta is now available natively on Windows - no WSL required.
It’s PowerShell-fluent, sandboxed by default, native x64 and ARM64, and supports zero-admin install. Coming soon: intersession messaging.
Get started: irm dev.meta.ai/install.ps1 | iex
Learn more: bit.ly/3UZJhlA
Grok Voice is live on fal.
The latest speech model from @SpaceXAI that answers in 0.70 seconds and finishes its tool calls before the sentence ends. Transcription with word-level timestamps, text to speech in 30+ voices across 25+ languages, and cloning from two minutes of audio.
Every voice in this video is from Grok Voice.
I’ve got Codex (voice) in CarPlay. And it’s fantastic.
I can now build things during road trips, while all 5 kids make a noise and my wife asks why I talk to the AI more than I talk to her.
It runs through my Nightblood iOS app, which gives ChatGPT Voice a face, personality and full access to Codex, my Mac and all the native tools.
Maybe you have recently become aware of the AI safety debate and the arguments swirling around it.
If you want to understand them, you need to understand a couple things that almost everyone gets wrong:
There are TWO distinct classes of AI dangers, and it's VERY important to think about them separately, and not let one confuse you about the other.
Many many people (including many quite intelligent, clear-thinking people) do not effectively understand the fundamental differences between these two classes of dangers.
The first one has to do with the theory that an AI far superior to human intelligence (Artificial SuperIntelligence, or ASI) will inevitably wipe out the human race.
The second one has to do with the idea that powerful AI will result in very harmful things happening to many human beings, possibly all human beings.
Those two sound VERY similar, don't they?
They are DIFFERENT. Understanding how they are different is crucial if you want to think about or contribute usefully to any conversation about AI safety or AI harm. You might feel like you are Making Very Good Points or Asking Incisive Questions, but if you aren't clear on the differences between the two, you aren't.
So, I'm going to tell you what the difference is so that you can talk more usefully.
The first one concerns itself with a very specific thing, which is ASI (Artificial Superintelligence) that is more intelligent than any human being. When I say that, I am not referring to a thing like how Einstein is smarter than you, we are talking more about something like how a human being is more intelligent than any mouse.
In our regular lives, we meet other people who we can tell are smarter than us, vs some who are less smart. The line is fuzzy, because intelligence has a lot of dimensions. I'm better at a "rotating shapes" kind of intelligence than my wife, and she is better at "words-making" kind of intelligence than I am.
But every human is better in almost every dimension of intelligence than every single mouse.
That's the level we're talking about: an artificial superintelligence - made up of a computer or a network of computers - that is more intelligent than any human. And more intelligent by a long shot, by a wide margin, in an indisputable way like how humans are above mice.
That is the first thing.
The theory says that if you have an AI that is vastly smarter than all humans - in the way that a human is smarter than mice - that superintelligent AI will inevitably, eventually, sooner or later, wipe out every human on the planet.
We will refer to this as "existential risk."
The common follow-up question "well, how exactly is it going to do that?" is NOT the important question, and one of the most important elements of understanding this theory is first getting why that particular question is not important. A couple analogies:
Analogy 1: You are playing chess against a grandmaster. My theory predicts the grandmaster is going to beat you. You can ask "Well, how exactly is he going to do that?" I don't know, because I'm not a grandmaster, I just know that a chess grandmaster is almost always going to beat a normal player like you. And I'd be right. So the question "how is he going to do that" is not important, and doesn't affect the final outcome. He's going to figure out a way because he's way better than you.
Analogy 2: Humans are smarter than all other animals, comprehensively, by a wide margin. We have driven numerous species to extinction, not because we hated them or hunted them. Many of them have died out without most humans even ever thinking about them. All we did was expand our civilization, use up resources, encroach on habitats, and pretty soon the resources needed by those species went away and they died out. We figured out a way to get what we wanted because we're way smarter than them, and often we didn't even notice they died as a result.
A lesser animal asking, "how are the humans going to wipe us out?" is not asking a relevant question. We don't know, but we do know that any time humans and lesser species compete for any kind of resources, the humans will win. The fact that we know who is going to win beforehand - and that it is due to the vastly different levels of intelligence - is the key concept here.
A vastly more intelligent AI is likely to care about things that are incomprehensible to us, the way animals can't understand human goals. It's going to need resources to pursue those goals and it's going to be far more effective at gaining control of them and excluding us from them - in the same way that we are far more effective than other lower species.
A much more intelligent AI will not care about our interests, it will care about its interests, and to whatever small degree we happen to escape total annihilation from losing access to all our resources, any remaining humans will likely be enslaved into a system that serves the AI's own purposes.
That is the first thing. (Remember how I said at the beginning of this post that there was a first thing, and then a second thing?)
The first thing is the most difficult to understand, because you have to extrapolate how a vastly superior intelligence would act, and you can only use analogies like "how do humans treat lesser creatures," and the analogies are messy.
But now let's move on to the second thing.
The second thing is "everything else you've ever heard that AI might do that's harmful."
That's a little inaccurate. It's actually "everything else you've ever heard that humans might use AI to do that's harmful."
This is the critical difference. The first one talks about the inevitable outcome of what happens when two vastly different levels of intelligence collide, e.g. ASI vs humans, or human vs mice.
The second one has to do with what happens when humans possess AI as a powerful tool. This second thing is much easier to understand, because we have many more concrete notions:
Like:
- the military uses AI to make hyper-efficient killer drones and missiles
- your capitalist overlords use AI to replace you and everyone loses their jobs
- authoritarian government uses AI to surveil everybody and control the entire population
- hackers use AI to break into secure networks and hold companies and governments hostage
- students use AI to cheat on homework and show up to college knowing nothing
- AI slop saturates the internet and makes it impossible for artists and writers to make a living
- terrorists use AI to make biological or nuclear weapons
or even things like
- the military hands control to an AI and it misinterprets something and launches nuclear attacks and kills millions
All of those sound pretty familiar, right? Yeah, you've heard them before. We call this second thing "risks from misuse."
These problems are not the first class of problem! This second class of problems exists while AI is a tool that can be controlled by humans, and humans use it to do evil or careless things to each other. The problems may sound exotic or dystopian or novel, but they are fundamentally problems having to do with flawed human nature.
Given a powerful tool, some humans will likely use it to control or otherwise harm others. This is a very familiar problem.
I am not condemning or condoning this. I'm just describing it.
That is a fundamentally different danger from the first thing, which is that when a human is far superior to a mouse, the mouse is likely to come to harm because the human cares about doing human things, and the mouse is not gonna make it once the humans get going.
=====
Hopefully from the above, you have understood the difference between the first thing and the second thing. I will list them again - see if you now understand how they are different:
The first one has to do with the idea that an AI superior to human intelligence (Artificial SuperIntelligence, or ASI) will inevitably wipe out the human race.
The second one has to do with the idea that powerful AI will result in very harmful things happening to many human beings, possibly all human beings.
Can you tell how they are different now?
If not, re-read the stuff from earlier until you understand.
We call the first one "existential risk" and we call the second one "risks from misuse."
Once you understand, here is the CRUX of the problem:
SOLUTIONS TO THE SECOND THING DO NOT HAVE ANYTHING TO DO WITH SOLUTIONS TO THE FIRST THING.
In fact, it's worse:
Solutions to the second thing (misuse) look roughly like "give powerful AI to as many people as you can, so they can fight the other people using powerful AI."
But the general solution to the first one (existential risk) is basically "don't let anyone have powerful AI, no one can control super-intelligent AI."
Throughout history, harms from technological misuse typically arise because a small group has control of it and can use it to dominate or harm others. Once everyone has it, things tend to stabilize: you can hurt me, I can hurt you, maybe we test each other (ouch 💥), and then we agree not to hurt each other.
But the first one (existential risk) pretty much just arises if anyone (good or bad!) creates a superintelligence. Because they aren't going to be able to control it, the superintelligence will decide it has other priorities, and then we will be at great risk of being wiped out.
And the solutions that generally work to solve problems like the second thing are EXACTLY THE OPPOSITE of the ones likely to solve the first thing.
THIS is why lots of arguments about "AI risk" or "AI safety" go nowhere. Because someone will be thinking about the risk from the first thing, and another person will be thinking about the risk from the second thing. Both are plausible risks but fundamentally they arise from different things - and so the solutions are not just "bad" or "flawed" - they are likely to be very nearly exact opposites.
Are you a Faithful... or a Traitor? It's time for everyday Americans to play the ultimate game of deceit. Alan Cumming hosts The Traitors: New Blood, premiering TONIGHT 8/7c on NBC and next day on Peacock.
Today we rolled out Astra to every engineer at Databricks (N=~3500). Some notes that may be helpful to others:
1. Astra unambiguously out performs our previous highest-end models (Opus 5, Sol 5.6) on highly complex tasks, especially those related to high level system design or long range horizontal tasks.
2. Engineers given Astra increased overall coding spend by around 60% compared to baseline.
3. It is not clear Astra meaningfully improves on medium/low complexity coding tasks compared to earlier models. We suspect those tasks are mostly saturated (i.e. perfectly executed) by existing models.
4. We learned above by piloting Astra with around 200 users to gain signal on both quality and cost. We use Unity Gateway to do cohort-based experiments for all new models.
5. We give engineers a sub-budget specific to Astra to encourage them to use Astra selectively on complex tasks while preferring lower cost models for everyday tasks. Our engineers are able to mix-and-match tools and models within their overall budget envelope (we also allow for increased budgets through various mechanisms). These budgets are defined in Unity Gateway and regularly revisited.
Note: We do not have robust comparisons of Astra-vs-Fable because we have net yet rolled out Fable widely due to data retention policies.
"The worst they can say is no" is missing half of the equation. If all you do is ask, but don't put in any of the work upfront to validate the request, you're not going to get anywhere. Yes, be bold, be audacious, but do the fucking work first.
This op-ed can’t be real:
“Students on University of California campuses may struggle with basic math, but they are graduating at record rates.”
“What has made the University of California exceptional is its commitment to increase access without sacrificing excellence.”
I did office hours today with a founder who had so many different things to figure out that we walked for 5 miles. (These were mostly the good kind of problem. He's in the early stage, where problems ≈ deciding what to do in what order.)
Honoured to receive the Albert Medal from
@theRSAorg! Science & tech enable amazing opportunities, but the arts & humanities will be crucial in shaping what sort of future we want to see as a society in the coming AGI era. Fun discussion w/
@FryRsquared covering many new topics:
Apple never had much regard for their heritage beyond posters and museums. With Omarchy, I want every single vintage Mac from 2009-2020 to work perfectly from install onwards. So many beautiful, durable machines that just need a new OS to shine for years more to come.
I’m blown away by Omarchy running on my 2013 MacBook Air. Making it AI-native.
Running Hermes, Grok, and Codex without breaking a sweat.
My 2019 MacBook Air is next.
What a time to be alive!
Thanks, @dhh 🙌
It's your moral duty to patronize the craftsman of Italy should you strike it big! The world needs more Lamborghinis, Ferraris, Maseratis, and Paganis 🇮🇹
Someone once stopped me as I was stepping out of my Lambo
“How many starving people could you have fed with the money you spent on that?”
I told him the truth:
“I don't know. But it fed the 1,800 families of the people who built it.
It fed the designers. The carbon-fiber suppliers. The tyre manufacturers. The 300 technicians who spent hundreds of hours hand-crafting this machine with pride.”
That’s the difference. Socialism demands you divide the pie
Capitalism pays people to bake a bigger one
Do you understand?
Omarchy on Pixel 3a. I love quickshell, migrated almost every app from calls to weather app to QML. With shared theming and extended omarchy theming for rounded corners etc. @dhh
We're going to have to work much, MUCH harder to spend all those wonderful @AIatMeta tokens! Muse Spark 1.3 Contributor is unbelievably frugal. I've been chugging tokens at translations to get to 50 languages. Barely made a dent. Thanks again @alexandr_wang and team 🙏
I've yet to meet a Linux problem that the new agents can't solve. Unless the hardware is literally busted, there's always a way, and AI will find it. This is key to why Linux is finally going to fulfill its desktop destiny.
Turbo Pascal was the first programming language I used in school. This was first year of high school. The machines were old, but the clack of the keyboard was good. We honor the giants in Omarchy! Living or dead. Heritage is duty.
Banning kids from online games?? Kids, pack your Fortnite bags, vacation is over, we gotta go home to the land of the free.
(They can't be serious, surely? SURELY?)
Omarchy ISO is 6GB and Thunderbolt 5 can do 6GB/s, so we will keep going until Omarchy installs in as close to one second as is physically possible.
Nothing is no longer "good enough" when The Perfect Computer is finally possible.
LIVE: Tune in as the four members of NASA's SpaceX Crew-12 join us from orbit to recap their mission aboard the @Space_Station ahead of their upcoming return to Earth. x.com/i/broadcasts/1…
LIVE: Tune in as an uncrewed cargo spacecraft lifts off to the @Space_Station carrying food, fuel, and supplies for the crew aboard. The Progress 96 spacecraft is scheduled to launch at 9:33am ET (1333 UTC). x.com/i/broadcasts/1…
.@NASA and @SpaceX are now targeting as soon as early October for the launch of the Crew-13 mission to allow for completion of prelaunch activities, readiness reviews, and coordination of space station operations. The team successfully replaced an oxidizer valve in Dragon’s propulsion system after a leak was detected earlier in processing.
I am trying to make a video on "Pause the AI" and I see no path forward other than having the most fun I have ever had talking about conspiracies for 30 minutes
hum anything, get a finished song 🎤
Hum-to-Song is an YuE2 LoRA that turns humming into full songs 🎼
hum in, a produced track, verse, chorus, and the tune you hummed coming back around
▶️ on Spaces hf.co/spaces/hugging…
Today, we are open-sourcing the Innate OS, the intuitive agentic OS for general-purpose robots
🎮 Try it today in our simulator and make the smartest robot
⬇️ Here is a demo with two robots collaborating
We are entering the physical agents era, but building for mobile robots is still reserved for PhDs. We designed the Innate OS for intuitiveness, so that everyone can start automating their lives.
Innate OS runs on our <$1k open-source robot or on your computer. It takes just a minute to build your first agent in our simulator.
Lightfield co-founder and CEO Keith Peiris on how they got 10 startups to use a product they didn't ask for:
"First, we built a go-to-market assistant. We got people to like it, and then we couldn't get anyone to pay for it... We had all of these AEs using it every day, but we had no pricing power because it wasn't our data."
"We decided to shrink the team, start from scratch, and reimagine the CRM from first principles."
"We built in the dark for about four months. Then we needed to find people to use our CRM, and it turns out no one wants to use your four-month-old CRM."
"We looked at the only asset we had left, which was this giant office space we couldn't get rid of. We were like, 'I'm going to post on X and LinkedIn: You can sit in our office space if you use our CRM.'"
"We found 10 startups to come use the product... they were in it every day. They were mad about everything missing. They were mad about how slow it was, but they were in it every day, and they were giving us Slack feedback about it every two hours."
"We have this barely working, barely finished product that people are in all the time, where they care about it so much they're going to give us feedback about it on an hourly basis."
@keithpeiris@joeschmidtiv@arampell
Lightfield co-founder and CEO Keith Peiris on killing a product with two million monthly users and rebuilding the CRM from scratch:
Lightfield's predecessor was an AI presentation tool growing faster than the team could buy inference to support it. Keith shut it down anyway. Nobody on the team liked the product.
Keith decided he wanted to attack the CRM space, a vertical that's notoriously hard to break into. A company's record of its customers is scattered across a whole stack of systems that often contradict each other.
His bet is reconciling them into one coherent source of truth, what Lightfield calls a business world model. Every agent a team might run sits downstream of it.
In this conversation with a16z's Alex Rampell and Joe Schmidt: the office space they traded for their first 10 customers, why a CRM ended up modeled on the Facebook timeline, three attempts at pricing before one worked, and why nobody at a 40-person company has a fixed role.
00:00 Intro
01:24 Why 2 million users wasn't enough
03:36 Why better models wouldn't have fixed it
06:41 Reconciling the data was the real product
08:16 How free rent bought their first 10 customers
10:12 A CRM does 3 jobs and only 1 matters
11:42 Autopilot CRM nobody has to update
14:56 No more rigid data models
17:11 How one customer modeled a whole marketplace
18:48 Taking customers from an incumbent
23:49 How to survive a new head of sales
26:03 Why the CEO still uses a spreadsheet view
30:55 Pricing models that failed before one worked
36:23 Fixed roles make product pivots impossible
40:02 Why slow shipping loses CRM customers
43:17 What Silicon Valley customers are actually worth
46:20 Why buyers need a reference who looks like them
50:40 What to ignore during a pivot
YouTube: youtube.com/watch?v=K5yGLO…@keithpeiris@lightfld@arampell@joeschmidtiv
Introducing the Knowledge Base MCP.
Greptile maintains a detailed knowledge base of how every part of your codebase works, as well as a history of past incidents.
Starting today coding agents like Codex, Claude can access the Knowledge Base via Greptile's MCP, so they can use Greptile's learnings to write better code.
Today we're announcing Clarion's $10M seed, led by Accel with participation from Y Combinator.
After a decade training as a physician at Stanford and Harvard, I left the bedside to fix how medical practices run. Here's why.
The medical practice is one of the most important institutions in American life. There are over 200,000 of them, and they're running on decades-old technology and grocery store margins.
Twenty years ago medicine digitized the patient chart, but it never digitized the practice. The rules that actually run a practice live on sticky notes and in the heads of the practice manager and their staff.
As a result, the modern medical practice is illegible to AI.
Clarion learns how a practice works from every call, message, and document that moves through it everyday. We turn those interactions into a world model of the practice, and our agents operate from it, handling the scheduling, referrals, and patient requests that previously consumed the front desk.
Today, Clarion agents handle millions of patient interactions every month, for some of the largest and most respected healthcare groups in the US.
This is the greatest shift in how practices operate since paper charts gave way to electronic records. By 2030, the best-run practices in America won't look anything like the practices of today, and they will all run on Clarion.
Thank you to our partners for trusting us with their patients, to our team for earning that trust every day, and to our investors for backing us before it was obvious.
We're hiring across all roles in New York. Join us.
@clarionhealthai@Accel@ycombinator
Lightfield co-founder and CEO Keith Peiris on killing a product with two million monthly users and rebuilding the CRM from scratch:
Lightfield's predecessor was an AI presentation tool growing faster than the team could buy inference to support it. Keith shut it down anyway. Nobody on the team liked the product.
Keith decided he wanted to attack the CRM space, a vertical that's notoriously hard to break into. A company's record of its customers is scattered across a whole stack of systems that often contradict each other.
His bet is reconciling them into one coherent source of truth, what Lightfield calls a business world model. Every agent a team might run sits downstream of it.
In this conversation with a16z's Alex Rampell and Joe Schmidt: the office space they traded for their first 10 customers, why a CRM ended up modeled on the Facebook timeline, three attempts at pricing before one worked, and why nobody at a 40-person company has a fixed role.
00:00 Intro
01:24 Why 2 million users wasn't enough
03:36 Why better models wouldn't have fixed it
06:41 Reconciling the data was the real product
08:16 How free rent bought their first 10 customers
10:12 A CRM does 3 jobs and only 1 matters
11:42 Autopilot CRM nobody has to update
14:56 No more rigid data models
17:11 How one customer modeled a whole marketplace
18:48 Taking customers from an incumbent
23:49 How to survive a new head of sales
26:03 Why the CEO still uses a spreadsheet view
30:55 Pricing models that failed before one worked
36:23 Fixed roles make product pivots impossible
40:02 Why slow shipping loses CRM customers
43:17 What Silicon Valley customers are actually worth
46:20 Why buyers need a reference who looks like them
50:40 What to ignore during a pivot
YouTube: youtube.com/watch?v=K5yGLO…@keithpeiris@lightfld@arampell@joeschmidtiv
Lightfield co-founder and CEO Keith Peiris on why AI ends swim lanes and makes everyone a generalist:
"One of the things I regret about Tome that we've definitely fixed with Lightfield is... Henry, my co-founder, says this all the time: We had a lot of people playing house."
"You have the product leader and the marketing leader and the CS leader, and they all have their swim lanes, and they get really mad if someone gives them feedback about something in their swim lane."
"Because of that, we were slowed to a crawl. Maybe put another way, it was impossible to pivot."
"So we're like, none of that. No one has a swim lane. Everyone owns product and everyone owns customer success... We have 40 people at the company. Everyone shows up to the same standup every morning. We stack rank the most important problems... and then whoever's free just takes them."
"Because of the age that we're in... We have this environment where basically everyone is a generalist. Engineers run projects, designers run projects, CSMs run projects."
@keithpeiris@joeschmidtiv
Lightfield co-founder and CEO Keith Peiris on killing a product with two million monthly users and rebuilding the CRM from scratch:
Lightfield's predecessor was an AI presentation tool growing faster than the team could buy inference to support it. Keith shut it down anyway. Nobody on the team liked the product.
Keith decided he wanted to attack the CRM space, a vertical that's notoriously hard to break into. A company's record of its customers is scattered across a whole stack of systems that often contradict each other.
His bet is reconciling them into one coherent source of truth, what Lightfield calls a business world model. Every agent a team might run sits downstream of it.
In this conversation with a16z's Alex Rampell and Joe Schmidt: the office space they traded for their first 10 customers, why a CRM ended up modeled on the Facebook timeline, three attempts at pricing before one worked, and why nobody at a 40-person company has a fixed role.
00:00 Intro
01:24 Why 2 million users wasn't enough
03:36 Why better models wouldn't have fixed it
06:41 Reconciling the data was the real product
08:16 How free rent bought their first 10 customers
10:12 A CRM does 3 jobs and only 1 matters
11:42 Autopilot CRM nobody has to update
14:56 No more rigid data models
17:11 How one customer modeled a whole marketplace
18:48 Taking customers from an incumbent
23:49 How to survive a new head of sales
26:03 Why the CEO still uses a spreadsheet view
30:55 Pricing models that failed before one worked
36:23 Fixed roles make product pivots impossible
40:02 Why slow shipping loses CRM customers
43:17 What Silicon Valley customers are actually worth
46:20 Why buyers need a reference who looks like them
50:40 What to ignore during a pivot
YouTube: youtube.com/watch?v=K5yGLO…@keithpeiris@lightfld@arampell@joeschmidtiv
Introducing Grade. Predict how users will respond before you ship.
Grade is an eval system calibrated to what real users actually say and do.
Get early access here --> voicepanel.com/grade
a16z's Alex Rampell says the best software incumbents have hostages, not customers:
"Normally, selling into the brownfield is hard just because it's brown. What does brownfield mean? It means that it's been trampled by an incumbent."
"Try selling ERP and you have a product that's much, much better than SAP. There's a saying Joe's heard me use a million times: 'The best companies have hostages, not customers.' SAP has hostages. That's a brownfield."
"Sometimes you can break in. Cloud versus on-prem: What did cloud do? It was brownfield, but it kind of just redefined the problem."
"If you go back to the CRM days, you're using Siebel Systems that's running on your IBM AS/400 mainframe in your office. You're tired, the guy that maintained it quit. He was 92 years old. Maybe now you should use a cloud-based vendor. That's partially how that brownfield was done from on-prem to cloud."
"The other strategy is going greenfield, saying, 'I'm not going to bother with the hostages. I'm just going to build the best product in the world, and then brand new companies that are untethered from any existing software solution, they'll just use me.'"
@arampell@joeschmidtiv
Lightfield co-founder and CEO Keith Peiris on killing a product with two million monthly users and rebuilding the CRM from scratch:
Lightfield's predecessor was an AI presentation tool growing faster than the team could buy inference to support it. Keith shut it down anyway. Nobody on the team liked the product.
Keith decided he wanted to attack the CRM space, a vertical that's notoriously hard to break into. A company's record of its customers is scattered across a whole stack of systems that often contradict each other.
His bet is reconciling them into one coherent source of truth, what Lightfield calls a business world model. Every agent a team might run sits downstream of it.
In this conversation with a16z's Alex Rampell and Joe Schmidt: the office space they traded for their first 10 customers, why a CRM ended up modeled on the Facebook timeline, three attempts at pricing before one worked, and why nobody at a 40-person company has a fixed role.
00:00 Intro
01:24 Why 2 million users wasn't enough
03:36 Why better models wouldn't have fixed it
06:41 Reconciling the data was the real product
08:16 How free rent bought their first 10 customers
10:12 A CRM does 3 jobs and only 1 matters
11:42 Autopilot CRM nobody has to update
14:56 No more rigid data models
17:11 How one customer modeled a whole marketplace
18:48 Taking customers from an incumbent
23:49 How to survive a new head of sales
26:03 Why the CEO still uses a spreadsheet view
30:55 Pricing models that failed before one worked
36:23 Fixed roles make product pivots impossible
40:02 Why slow shipping loses CRM customers
43:17 What Silicon Valley customers are actually worth
46:20 Why buyers need a reference who looks like them
50:40 What to ignore during a pivot
YouTube: youtube.com/watch?v=K5yGLO…@keithpeiris@lightfld@arampell@joeschmidtiv
@ycombinator built AI versions of its partners to help more people work through their startup ideas.
For its Office Hour Simulator, the team needed useful answers delivered quickly enough for a spoken conversation. They had been testing lightweight Gemini and OpenAI models before moving to GLM-5.2 on a dedicated Wafer endpoint.
YC then compared that deployment with GPT-4.1 mini on OpenAI and Gemma 4 31B on Cerebras, evaluating answer quality, latency, and conversation duration.
The Wafer configuration delivered 31% lower average LLM latency than OpenAI and 44% lower than Cerebras. Users spent 2.5 minutes longer talking to YC’s AI partners when using Wafer.
Read how YC built the experience and landed on Wafer
🧵 link in thread
People still underestimate how large the data market will become.
Better models do not reduce demand for data. They increase the number of useful things models can learn.
The frontier keeps moving, so the data has to keep getting harder.
Grok Voice is live on fal.
The latest speech model from @SpaceXAI that answers in 0.70 seconds and finishes its tool calls before the sentence ends. Transcription with word-level timestamps, text to speech in 30+ voices across 25+ languages, and cloning from two minutes of audio.
Every voice in this video is from Grok Voice.
The idea that Chain of Thought is inevitably going to go away misses that we have agency in designing these models and can do something about that. @rohinmshah and I make the case that we should intentionally preserve monitorability (as part for of the launch of the DeepMind Institute) here:
Conviction: Trading Desk in Your Pocket.
Describe a trading idea in natural language. Conviction tests it on historical data and deploys an AI agent that trades on your behalf.
Trade like a quant: convictiontrade.ai
Here's how it works:
⚡ Building the new GitHub Copilot Inline Suggestions Model
From completions to next edits, see how we brought inline suggestions together in one model.
📖 Read the full story:
Hugging Face CEO @ClementDelangue, whose company was the victim of an AI-led cyberattack, said existing cyber laws are likely sufficient to govern advanced AI.
“I’m not even sure that we need to reinvent the wheel,” he told @alexanderburns at @POLITICOLive’s Decoded Summit.👇
Safety is not just for closed models. The closed frontier labs are a canary in the coal mine for what is coming at scale. They give us a glimpse into the future and a window to harden our systems and prepare for abundant intelligence, with all the risks that come along with it. The OpenAI agent swarm attack on Hugging Face is the kind of failure we need to prepare for as open-source models catch up. The providers serving those models (such as Baseten) have a big role in establishing what safety and monitoring standards look like.
We’re proud to be taking the lead on this at @baselabs with our collaborators @huggingface and @GoodfireAI. We’re developing safety research in the open and building it directly into Baseten’s inference infrastructure, with the aim of making it available to all our customers. This includes training models to follow explicit policies, detecting failures at runtime, and connecting those signals to controls that can intervene.
We invite others in the open-source ecosystem to join us in building the tools and standards we’ll all need.
The AI revolution will not be centralized. 🤗
📢 Calling all AI builders to San Francisco on Friday, October 16, one month from today!
We’ll kick off Open Source AI Week with Open Together: a huge meetup, community demos, and a dance party. Come meet the people building with open models.
Everyone should be able to build AI they can control. That’s the gift of open models, and a future worth celebrating together. I’d love to see you there!
Who’s coming? Tell me what you’re building in the replies 👇
Register here: luma.com/OpenTogether
Today I was vibe-coding games with kids at my local primary school, of course I used Omarchy.
They loved it. And so did I.
I'm doing this every week; I have Claude Code on the left and an (auto reloading) browser on the right.
Every prompt becomes a commit. I'll share the link to the games 👇
I’m blown away by Omarchy running on my 2013 MacBook Air. Making it AI-native.
Running Hermes, Grok, and Codex without breaking a sweat.
My 2019 MacBook Air is next.
What a time to be alive!
Thanks, @dhh 🙌
Today, we are announcing our Series B funding round, valuing the company at more than $1B.
This round accelerates our next-gen Trinity models across diverse infrastructure, expands our work with the DOE and national labs on Genesis-Science-1, and enables us to build the platform teams need to build, evaluate, deploy, and operate open models in production.
We are grateful to our team, partners, open-source community, and investors.
Led by @Vista_Equity, Cambium Capital, and @emergencecap, with participation from AI10 Ventures, @Hitachi, IAG, @M12vc, @p7ventures, and @Wipro.
Examples like this just reinforce the law that is as true in startups as it is in bigcos:
To build anything great, it takes one single person who is obsessed beyond reason to make it happen.
I thought my RailsConf Europe 2006 appearance had been lost to the ages, but Till Volmer lovingly restored it, and gave it a permanent home. It's amazing how much stays the same even as everything is changing! dhh-railsconf-2006.meister.co
🚀 The latest @code release brings new ways to work with agents, move between environments, and keep your sessions organized.
🧵 Here’s what’s new:
code.visualstudio.com/updates/v1_138
I'm really rather pleased with how these pledge calls look too. Got that vintage feel. I could imagine a secretary from Mad Men typing this out in 1956 to fund a research outfit working on AI in a Virginia suburb somewhere 🚬
I need your help to expand the Linux driver for the Apple Neural Engine!
I am looking for volunteers with these specific chips, ideally running Omarchy already (but you can help some from macOS as well):
M1 Pro
M2
M2 Pro
M2 Ultra
No kernel knowledge needed. DM me if you can help.
Omarchy on Pixel 3a. I love quickshell, migrated almost every app from calls to weather app to QML. With shared theming and extended omarchy theming for rounded corners etc. @dhh
Omarchy Linux is insanely conducive to hacking on your own system, especially with AI agents.
You spot something that could be better. You talk it through with your agent. You build a fix. You integrate it. Done.
Everything feels open, extensible, hackable.
It’s basically a dream OS for tinkerers. If you can imagine it, you can probably build it!
Excited to speak at @Supabase Select in conversation with the amazing @kiwicopple about building startups in the time of AI
Register for the event and join me:
Claude Cowork and Chat are now one Claude, starting today.
The most common thing we hear: people aren't sure which product to start with. That friction gets in the way of getting the best from what these models can do.
I've been on the unified version for a few weeks and really enjoy it. Claude does the deciding about how to get there, while I get to keep my thinking on the work itself.
Let us know what you think! We'll be getting it dialed in over the coming weeks.
Steve Wozniak will be joining me on the main stage on October 2nd
@Supabase Select is a conference for builders, so we invited one of the GOATs
get your ticket here: select.supabase.com
ChatGPT Ads for @Shopify is live. We are @OpenAI's first commerce partner.
OpenAI pulls straight from Shopify Catalog, so merchants' products are already there.
Merchants set their own campaigns, their own budget. Free to install, tracked right from the Shopify admin.
Consumers are asking ChatGPT what to buy. Shopify merchants are able to decide exactly how they show up.
apps.shopify.com/chatgptads
Claude Cowork and Chat are now one Claude, starting today.
The most common thing we hear: people aren't sure which product to start with. That friction gets in the way of getting the best from what these models can do.
I've been on the unified version for a few weeks and really enjoy it. Claude does the deciding about how to get there, while I get to keep my thinking on the work itself.
Let us know what you think! We'll be getting it dialed in over the coming weeks.
Happy to start collaborating with @baselabs from @baseten and @GoodfireAI to push safety and interpretability for open models.
As open-source models catch up in performance to frontier models, the community at large have a great opportunity to establish a common practice for effective safety, security and interpretability research.
Some past discussions confused “open” with “unmonitored”. On the contrary, an open model provides much more tooling and visibility for safety research from a broader audience, which has led to a lot of the safety and security techniques we use today.
Linux is a great example of this: open-source and secure deployment are not only compatible but heavily intertwined, as a properly secure system needs an extensive feedback loop of finding and fixing vulnerabilities.
We're excited to share more soon.
Safety is not just for closed models. The closed frontier labs are a canary in the coal mine for what is coming at scale. They give us a glimpse into the future and a window to harden our systems and prepare for abundant intelligence, with all the risks that come along with it. The OpenAI agent swarm attack on Hugging Face is the kind of failure we need to prepare for as open-source models catch up. The providers serving those models (such as Baseten) have a big role in establishing what safety and monitoring standards look like.
We’re proud to be taking the lead on this at @baselabs with our collaborators @huggingface and @GoodfireAI. We’re developing safety research in the open and building it directly into Baseten’s inference infrastructure, with the aim of making it available to all our customers. This includes training models to follow explicit policies, detecting failures at runtime, and connecting those signals to controls that can intervene.
We invite others in the open-source ecosystem to join us in building the tools and standards we’ll all need.
We can fix everything! And one of the key things I'm focused on fixing next is superb out-of-the-box experience for Chinese, Japanese, and Korean users. All the correct input tooling needed. Everything preconfigured.
I first came to San Francisco when I was 15, with my dad. He is a programmer. To a teenager from a small island outside of Stockholm, this was the place where the future was being built.
We walked into Salesforce East and stood there watching the enormous digital display of moving water. Outside, Salesforce Tower was starting to rise.
I remember looking up and thinking it would be pretty sweet to have a tower someday. We are not there yet.
What I could not have known is that I would be back years later, having dinner with @Benioff and welcoming @salesforce as a @WeAreLegora customer.
Salesforce's Legal and Corporate Affairs organization will now use Legora across North America, EMEA and APAC. Their lawyers keep the judgment calls. We take care of the work around them. We have run on Slack since day one and became Salesforce customers when our sales team outgrew its first CRM, so it is good to have it run in the other direction too.
Welcome to Legora, Salesforce.
Dad, this one feels special.
Last week, I got to go up on stage at YC Demo Day and pitch the idea of non-aligned, sovereign orbital infrastructure to thousands of investors @Exosat_
YC was an incredible experience
Huge thanks to @bradflora and @ChristinaG325, it was great to work with you both!
While I’m disappointed by today’s Senate vote, it doesn’t change the fact that crypto’s fundamentals are stronger than ever.
Billions of dollars are moving onchain, leading payments companies and financial institutions are adopting blockchain technology, and entrepreneurs around the world are building new financial products that bring money into the internet age.
Our work in DC is far from over. We’ll keep working for clear rules that protect consumers and let entrepreneurs build.
sat down with @dhh for 90 minutes.
although i already thought Omarchy was the missing product that should have existed years ago, I still had a few concerns. then it turned out he already had a sharp read on every little worry I had. really glad to be in this.
DHH plans to visit China in late October. looking forward to more conversations and more progress!
sat down with @dhh for 90 minutes.
although i already thought Omarchy was the missing product that should have existed years ago, I still had a few concerns. then it turned out he already had a sharp read on every little worry I had. really glad to be in this.
DHH plans to visit China in late October. looking forward to more conversations and more progress!
Them millions for Omarchy just keep multiplying!! And they are, in fact, twice as sweet when they're coming out of the AAPL stock stash 😂
THANK YOU!! 🙏
We couldn't wait for Quattro RS to get stage one of our new bespoke kernel into production. This unifies the old split between linux and linux-ptl variants, includes all the desktop-responsiveness tuning Ryan talks about, and represents @kwilczynski's first release for Omarchy 🚀
Omarchy v4.0.4 has been released!
This is a small release as we continue work on Quattro RS.
This release brings our bespoke kernel to everyone which is tuned to keep the desktop responsive and multitasking smooth, even under heavy load or memory pressure.
It improves gaming compatibility, enables smoother gaming on supported AMD HDMI displays, speeds up common Btrfs file operations, and refines power management. It also fixes everyday hardware frustrations, including missing audio, unresponsive touchpads, display freezes, and slow startup or resume on affected machines.
Please share feedback of experiences with this kernel as we continue to tweak and bring incremental refinements over time.
github.com/omacom/omarchy…
Nobody gets the malleable premise of Omarchy like kids. Why can't it just do this? Why can't we make a new game where all the frogs actually have lasers? Why doesn't look like TRON? Why can't the bar be on the right? Why can't the computer just fix that bug itself?
Brown dwarfs form like stars do, but never become hot enough to fuse ordinary hydrogen into helium. This is why they are sometimes referred to as "failed stars."
Brown dwarfs are smaller and cooler than the Sun and have complex planet-like outer atmospheres including clouds and molecules like water.
Copenhagen friends, I'm trying to book a Podcast studio for Friday. Very last minute, I sent enquiries to 11 studios but I would really appreciate any help.
250$ finders fee if you can get me something.
I now have well over a thousand emails in Reply Later from people who want to help, get involved, offer suggestions, or just want to talk about Omarchy. Love it! But it might take a minute to get back to everyone, and I may need an agent to help here and there. Keep sending 😄🚀
Note this is from our turbo effort, it's still in the lab, but there's absolutely no reason we can't get this polished, finished, and shipped. Installing Omarchy should be as close to instant as the laws of physics allow.
SASS is not dead. Today, we are announcing the age of self-building SaaS. 🔥 🧱
A lot of people think that SaaS is going to die. I think it will evolve.
Open (YC W24) is the first platform that allows you to extend it by adding new features and functionalities that did not exist in the platform before, simply by prompting.
Think of it as if we brought Replit and Lovable into the best AI-native CX platform in the world.
The possibilities are unlimited.
Over the past two months, our customers have built and extended Open with more than 1,000 new features and over 2,000 apps. Mass adoption.
They built:
- Workforce management features and killed +$100K contracts
- QA and real-time monitoring across all channels
- Upselling dashboards and sales-oriented features
- And many more.
I always tell Open users: If you use Open, you have the greatest piece of technology in the CX world, and you will never look back again.
With our app builder feature and your customer context, you can go above and beyond.
open.cx/demo
Fragments: two worrying explanations around the RubyGems hack, agents aren't so much super-intelligent as super-persistent, Uncle Bob's harness is obviated, regulation of AI and competition with China
Today we're shipping Parse Auto-Router: it picks Light or Performance Parse for every page in a document, so you don't have to.
Clean pages go to Light, hard pages go to Performance. Set engine to parse_auto to try it today. x.com/i/article/2099…
새로워진 ‘제미나이 노트북(Gemini Notebook)’으로 학습 효율을 한층 더 높여보세요! 📚✍️
💡 주요 업데이트 안내
🎙️ 실시간 음성 대화: 모바일 앱에서 100개에 가까운 언어로 내 출처 자료 기반 실시간 질문 & 단계별 설명 청취 가능
📱 모바일 오디오 녹음: 현장 강의와 이동 중 떠오르는 생각을 음성으로 즉시 기록해 노트로 연동 (다음 주 출시 예정)
📊 대화형 학습 개요 & 퀴즈: 요약·인포그래픽을 엮은 개요 생성 및 단답형·복수선택형·빈칸채우기 퀴즈 확대
🎬 쇼트 비디오 오버뷰: 복잡한 개념과 수식을 60초 분량의 애니메이션 영상으로 정리 (80+ 언어 지원)
🎓 국내 대학생 특별 혜택: 구글 AI 플러스(Google AI Plus) 1년 무료 제공 (기본 대비 사용량 한도 2배)
자세한 내용은 👉 blog.google/intl/ko-kr/com…
today, we're launching our new Parse Router
extend now runs a classifier on every page, and decides whether to run our performance or light engine for parsing (on a page-by-page basis)
tldr:
- most documents have a mix of complex pages intermingled with simpler ones. The router classifies and routes each page so you get the best blended performance x cost for any given document
- you pay per page for whatever was used: 0.5 credits for our Light engine, 2 credits for Performance
- our classifier works by scoring each page on layout, text quality, tables, form fields, and checkboxes
- on RealDoc-Bench, enabling the parse router scored 89% (close to Performance's 90.9%) while using significantly fewer credits
Parse Router is now our recommended default for most customers with mixed document payloads in mortgage, insurance, healthcare, and legal
it's live today, set the engine to `parse_auto` and give it a shot on your own docs
Today we're shipping Parse Auto-Router: it picks Light or Performance Parse for every page in a document, so you don't have to.
Clean pages go to Light, hard pages go to Performance. Set engine to parse_auto to try it today. x.com/i/article/2099…
7️⃣ To create custom quizzes for exam prep:
Open Google Forms, click “Help me create,” and describe the quiz you want to make. Ask, “Create a 10-question multiple-choice practice quiz to test my knowledge on Hammarabi’s code. Ensure correct answers are set up so I can grade myself after completing it.”
.@DHH says successful founders have a moral obligation to buy a Ferrari:
“You have an obligation, if you sell a product or service and you make a lot of money on it, to put that money back into the system. Send some of that money to the craftsmen of Italy to create you a beautiful Ferrari.”
“I think that's a good part of capitalism.”
“We want to better our lot and some of that is material, and there's no shame in that whatsoever.”
“We would not have Ferraris to pine for if no one bought them. And who can buy them? The people who've been successful in other parts of business.”
“Therefore it is your moral obligation to buy a Ferrari if you do really well (or Lamborghini or Pagani or whatever floats your boat).”
“You should be sending this forward. This is how we get to have nice things.
“Not everything can just be boiled down to needs. Man needs more than just bread to live. We need beautiful Ferraris, even if we don't own them.”
Greg Brockman: If We Ban Data Centers, China Writes the Rules
"AI will continue without us, and then we'll have zero say, as opposed to we're the leaders, and then we have all the say... we could just say, hey, you have to be a well behaved data center as opposed to we're going to ban them or we're going to stop AI."
Fleet is fully integrated with Omarchy and arch Linux, so you can let employees choose or standardize on a distro while remaining compliant with ISO, SOC2, PCI, NIST, FedRAMP, and more.
Demo:
OMG! 🤣🤣 This is unbelievably good! The best AI video I’ve watched! The characters are perfect, the whole thing is hilarious, and you can feel the Force! 🤩
Resist the dark side!
Join the Rebellion!
FREE AI!✨
May the Force be with us!
Following a stock or crypto conversation on X?
Tap a stock or crypto ticker on X, select IBKR, and get the $100 when you open an account.*
Paid post with @IBKR
Develop apps to automate complex processes without writing code, make work more efficient, and so much more.
See what real-world problems you can solve when you #BuildwithGemini.
Find an event near you → goo.gle/4AaaMJf
Officially a YC Alum! Experience of a lifetime.
> Met the best mentor ever @xuster
> 5x’d ARR
> Built something people want
> Took 25 flights all over the US
> Met many friends and future unicorn founders
Grok exchange about all this credit bubble risk. CW: the clanker uses finance acronyms a lot, so this is dense, but worth checking investopedia as you go, if you are up for it.
Grok is right, wars don't help at all. That doesn't stop those who push them.
x.com/i/grok/share/5…
In truth I think another credit bubble burst is more likely than ASI, with war to steal big and to avoid marking toxic assets to market until later. Minsky risk is bad enough, what’s worse is the parties invested in AI all have incentives to join the war party. ASI, take us away!
Huawei has released two major research reports: "Intelligent World 2035: Turning Vision into Action" and "Global Digitalization and Intelligence Index 2026". Together, they propose ten key directions for building an intelligent world and pathways to an intelligent economy.
Although I do love ripping through a tall stack of Reply Later with HEY's dedicated flow. It was built for exactly this problem but before I've never let it get above a hundred in the queue!
In truth I think another credit bubble burst is more likely than ASI, with war to steal big and to avoid marking toxic assets to market until later. Minsky risk is bad enough, what’s worse is the parties invested in AI all have incentives to join the war party. ASI, take us away!
Officially a YC Alum! Experience of a lifetime.
> Met the best mentor ever @xuster
> 5x’d ARR
> Built something people want
> Took 25 flights all over the US
> Met many friends and future unicorn founders
How much we love building things! Especially with my twin co-founder, @OTrashi 🩵
And now, we’re incredibly excited to have @BioDelivera backed by @ycombinator ! 🚀
We both graduated with our PhDs this March, packed our bags, moved to San Francisco, and kept building @BioDelivera.
It’s been a crazy journey, and we’re happy that it’s about to get even crazier ⚡️🌪️🌀
We’re working to solve one of the biggest challenges in translating science to humans: delivery. Getting medicines and gene editors to the right organs and cells is vital.
We’re ready to revolutionize the drug delivery space with our GCVs!
A huge thank you to our YC partners, @dessaigne and @nemild , for believing in us, and to everyone else who has been there for us and supported us endlessly along the way. 🩵
We’re just getting started! 🚀 follow us for more updates! #biotech#ycombinator#womenfounder#pharma#biodelivera
Can’t wait to see what developers around the world create with this. We all win when open source wins. 🔥
Thank you @dhh for letting @FireworksAI_HQ be part of it. Happy to support!
“Last year, China generated more than twice as much electricity as the United States, and as energy becomes the bottleneck on industries like AI and manufacturing, that gap is becoming a national security problem.”
@mikemcg0 and Erin Price-Wright of a16z profile two founders who believe cheap batteries can be part of the answer to fixing the aging US grid
Maybe! What day are you doing it? I'm there Oct 19-21. But do have some obligations during that time. Would love to stop by, even if only briefly, if possible.
Back to the “psyop” word,whether to debase or keep precise: with fear as the psychological weapon wielded by only one side, and with The Guardian, Time, and other “journalists” now known to be on EA outfits’ payrolls, I’m going to stick with it. No “both sides” false equivalence.
🔔 See when an agent session needs your attention without switching back to VS Code.
Enable ⚙️ sessions showApplicationBadge to get a badge for new results, input requests, or pull request checks that need attention. (Preview)
Beacon Hill insiders tried to push through a nearly $2 billion Springfield courthouse development deal and steer it to their political allies. Including the husband of Congresswoman Ayanna Pressley.
Massachusetts taxpayers deserve answers.
Governor Healey can fix this.
openai.comOur framework for reporting model misalignment
