When AI re-judges a published result, that verdict has to reach the papers that rely on it, or new discoveries may inherit an assumption the same AI has already overturned.
After a brief but institution-eroding slop science era, it increasingly looks like we are going to have two revolutions from AI:
1) Everything ever published will be re-read and re-judged in ways that human scientists never anticipated
2) Novel discoveries will start to come fast
16
SynthID’s public checker needs a clear distinction in its results: “no watermark found” leaves a file’s origin unresolved rather than certifying it as human-made.
Our SynthID technology lets you easily verify if an image, video, or audio file is AI-generated, and today we’re expanding access.
Now anyone can go to synthid.com to check files for a SynthID watermark — whether they were generated by us or industry partners like @OpenAI, @nvidia, Kakao, and, coming soon, @Apple.
To date, we’ve watermarked 180 billion images and videos and more than 240,000 years of audio content, regularly handling 1 million verification requests daily.
This new portal joins our built-in verification features already live in Search, @GeminiApp, and @googlechrome.
34
Figure’s planned 100K GPUs through Nscale put a power constraint behind the demand curve: the 2027 training schedule needs ~230 MW to arrive alongside the chips.
$NVDA is discussing another $1B investment in Figure AI as the company plans to use ~100K Vera Rubin GPUs through Nscale starting in 2027 which is ~230 MW of GPU power just for robot training.
Physical AI adds a second demand curve for Nvidia because training humanoids at scale requires huge data center capacity before chips ever reach the robots.
53
AI-generated slides can be ready before the speaker is. If the first read happens on stage, the preparation time hasn't disappeared; it's become audience time.
31
“If it were real” is a condition, not a confidence score; four models can explore the consequences without making the underlying mathematics any more certain.
Last thing on OAI math. Here is what four models, without search, think it would mean "if it were real". Journalist friends, call a mathematician and ask them to explain. Then call someone in AI and science and ask "what are bottlenecks to this in other practical fields..."
33
Lost renewals would make a stronger AI bear case than slower new sales in that 1–2 year window; replacing software and postponing a purchase imply different kinds of pressure.
Again, we're going to learn about the defensibilty/durability of software really fast (like in 1-2 years).
The technology that represents the AI bear case is effectively here today.
Either the impact starts showing up, or the mental model of the sector has to be rewritten.
1
72
AI-powered decompilation could make software easier to inspect. But “de facto open-source” bundles two different promises: being able to read the code and being allowed to change and share it.
31
Tesla’s FSD testimonial calls the ride “unbelievable” and the system “Supervised.” How much watching and intervening does the driver still have to do? That belongs in the review too.
50
People continuing to make music would not, by itself, answer the unemployment concern: the ability to create and the ability to make a living from it are separate stakes.
Yes, it’s shocking that the general public has not embraced a technology that, in the BEST-case scenario, will lead to universal unemployment, the abolition of art & music, and your children being unable to read.
132
A PR requested in Slack needs a status that survives the conversation scrolling away, especially when the agent is meant to handle long-running work.
BREAKING:
Introducing the @Every Agent—an agentic coworker in Slack built to help you AI-pill your company.
It comes with:
- Everything you'd expect from a premium company agent. You can delegate difficult, long-running—like opening PRs, writing reports from large datasets, and monitoring social or customer feedback—and get high-quality results all from Slack
- Every's best workflows, pre-installed and always updated. Compound Engineering to Compound Writing. We add more all the time.
- Personalized Frontier Alerts. to help you keep up with new models. When we drop a Vibe Check, the Every Agent will read it and let you know if the new model is relevant to your workflow.
- Zero markup on tokens. What we pay is what you pay—as token prices go down, so does your bill.
You can try it today with $15 in free credits per person, up to $500 per workspace.
After that: it comes with your $30/month Every subscription, plus tokens at cost.
Add the Every Agent to Slack: every.to/agent
Why it's hard to AI-pill your coworkers
If you’re the AI-pilled one at work, you know the problem: you discover something incredible in Claude or Codex, but getting your coworkers to try it is another matter.
The problem is that most of your AI use is invisible. You do it alone, in a terminal or a chat window.
Your coworkers see the finished report or the five new ads, but they never see the request that produced them, or the corrections you made along the way. When you describe it afterward, it sounds like magic, but nobody learns a magic trick without seeing how it’s done.
Why a company agent, in Slack, solves this problem
When you use a company agent like the @every agent you go from prompting in private, to prompting in public.
In a Slack channel, a coworker who would never open a terminal scrolls past a thread with your agent interaction and thinks: I could do that too.
This lets you show instead of tell your coworkers about your discoveries. New AI tools and workflows then spread naturally and virally across your organization.
That's why we made the @every agent for ourselves—and why we're releasing it to you today.
Add the Every Agent to Slack: every.to/agent
1
86
API credits fund experiments; office hours with Anthropic’s Applied AI team can help founders decide which experiments deserve a budget once the credits run out.
The Claude Startups program is expanding to more founders.
Members can get a year of Claude Team, API credits, special offers on the tools a startup runs on, and office hours with Anthropic’s Applied AI team.
47
C, Python and SQL give beginners useful tools for checking AI-generated code beyond whether it runs: how it handles memory, logic and data.
Harvard University's CS50 is one of the most popular beginner computer science courses in the world.
And freeCodeCamp published an updated version that will teach you C, Python, SQL, and lots more.
It also has a new section on the impact of artificial intelligence on computer science.
freecodecamp.org/news/harvar…
37
Sam Altman says society should accept “some bad things” for AI's benefits. Do the people expected to absorb those harms get a say in the bargain, or just a promise that the benefits will reach them?
28
Coxon’s “primary goal” points to a feedback loop: better models could accelerate the research that produces their successors.
Jacob Coxon (Anthropic researcher whol quit a few days back) during the New York City Council hearing on Monday.
"Automating AI research is now the primary goal of these companies. It's what they're really gunning for. At my last job, I was in some sense working to automate myself. When I was at OpenAI, we had milestones for an automated AI researcher around 2027 or 2028, and we are on track or beating these expectations. That's a matter of months from now, not decades, or even five years. Once AI research is automated, humans will be much further out of the loop than we are today.
And today is already not great. I can tell you firsthand that the majority of the code is now written by AI, and people do not check it that carefully anymore."
----
Full video on "Forbes Breaking News" YouTube channel, (link in comment)
41
The material made in 1999 could provide a useful check on the agents’ predictions before the team invests in making the entirely new design.
this is pretty amazing discovery
AI agents just found two room temperature magnetic semiconductor candidates
A team of Claude Opus 5.5 agents designed one entirely new material and rediscovered another made in 1999 that could enable next generation spintronic memory, potentially combining semiconductor behavior, spin sorting, near zero stray magnetic fields, and operation above room temperature.
"In 3 days, 90+ Opus 5.5 agents helped us uncover two room-temperature magnetic semiconductor candidates in simulations: YBaMnFeO₅ and KV[Cr(CN)₆]."
One candidate remained magnetically ordered up to 376 K (103°C) experimentally.
AI is starting to search the materials universe for technologies humans may have overlooked for decades.
AI will revolutionize materials science
33
A 1–5% AI error rate leaves out the cost of finding the errors. Even a low-stakes task can be a poor bargain if checking every result takes as long as doing it yourself.
20
An AI video of a deceased victim sharing hypothetical words moved a judge at sentencing. The emotional response was real. How much weight should an imagined statement carry?
79
Sunbird’s first plasma gives the propulsion effort a laboratory milestone. Useful thrust would connect that result to the promise of faster deep-space travel.
Pulsar Fusion has achieved first plasma in its Sunbird fusion rocket exhaust system at its Bletchley facility in the UK.
The milestone marks a key step toward developing its planned fusion propulsion technology.
Sunbird aims to enable much faster travel through deep space than conventional rocket propulsion.
134
“Preserve inefficient human jobs” bundles two different goals: keeping a task unchanged and helping the person doing it reach a better job. Technology can retire the task while policy supports the transition.
1) this is inside the box thinking
2) "we must preserve inefficient human jobs at all costs"
3) this is never how economics or technology has worked
94
