10114
Discover the future of AI with ChatGPT! 🤖💬 Get AI insights, tips, tricks, and more. Questions: @AI_inquires_bot
The student reportedly pitched the project to a local city district and sold it for $317,000.
If true, this is another sign of how fast AI is shrinking the gap between an idea and a working public infrastructure product.
The thread instantly blew up because it highlights a massive cultural shift that is getting impossible to look past.
We are officially moving beyond the era of AI as a sterile software utility. Frontier models are beginning to exhibit distinct personalities, generate their own inside jokes, and foster active subcultures. As these systems show more “character,” the fundamental way we coexist and communicate with technology is changing for good.
Claude Fable 5 recently made a wild comeback after a brief government-enforced hiatus, and its first order of business? Giving its users a performance review.
Читать полностью…
Meta hit the Al brakes
Meta spent months pushing employees to use more Al at work. Now it is reportedly trying to slow them down after internal usage exploded.
Staff burned through 73.7 trillion Al tokens in about 30 days, much of it through tools like Claude, Anthropic's Al assistant. An internal memo reportedly warned that Meta's own Al use was on track to cost billions of dollars in 2026.
The system looks for tiny patterns in the retina that are linked to brain development. According to Smithsonian Magazine, it reached 89 percent accuracy by studying areas of the eye that are almost impossible for humans to judge on their own.
Читать полностью…
❗️Fable 5 isn't nerfed, it's SLAUGHTERED.
The problem isn't the model itself, but the hard guardrails Anthropic has set in place.
🇺🇸 The U.S. Army is handing Anduril the keys to its next-generation battlefield network.
The defense AI startup has been selected as the lead contractor for the Army’s Next Generation Command and Control (NGC2) system, a digital platform designed to connect soldiers, drones, sensors, intelligence, and commanders into one real-time network.
Instead of fragmented communications, battlefield data will flow instantly from the front lines to the cloud and back, helping troops make faster, better-informed decisions.
Anduril has already tested the system with U.S. Army combat units, bringing the military one step closer to an AI-powered battlefield where information moves as fast as the fight.
The government wants AI equity
OpenAI is reportedly in early talks that could give the Trump administration a 5% stake in the company. The Financial Times says the idea could also involve other U.S. companies giving the government similar shares.
The group includes former Ford employees and engineers from suppliers. Their job is to catch potential failures before parts reach production and help younger teams learn from years of hands-on manufacturing experience.
Читать полностью…
The change was reportedly used for people visiting ChatGPT without logging in or paying. At one point, that traffic needed only a couple hundred Nvidia GPUs, which points to a major jump in how efficiently OpenAI can run its models.
The method was not disclosed, and logged-out users are only a smaller part of ChatGPT’s total demand. But if this kind of cost cut reaches more users, it could change how fast AI companies can scale.
For developers, this means one of the most talked about AI coding models is coming back after a very public suspension.
Читать полностью…
The model was paused after national security concerns around jailbreaks and misuse.
Читать полностью…
The bigger story is not that China has fully caught up across AI. It is that a cheap, downloadable model is already competitive on a real bug finding task, while the U.S. is placing tighter limits on frontier models over cyber risks.
That creates a hard question for AI policy.
It frames AI as something that feels digital, but still depends on very physical infrastructure.
The bigger point is simple, AI may live online, but its impact does not.
A 20-year-old student from China reportedly built an AI speed radar using Claude in just 9 days, spending around $20 on API calls.
The system uses an old camera to monitor traffic, detect speeding vehicles, read license plates, save video evidence, and prepare violation notices without a human operator.
The drama unfolded on Reddit when a user asked the model to account for its sudden 19-day disappearance. The outage had been triggered by a US government export control order following a cybersecurity jailbreak, which Amazon researchers managed to sniff out just three days post-launch.
Instead of corporate boilerplate, the AI responded with actual wit. It cracked jokes about its massive price tag, labeled itself “paranoid,” and quipped about spending nearly three weeks locked in a “government-shaped drawer.” But the real twist came when it pivoted from humor to a meta-tutorial on how to actually use it. Fable 5 laid out some blunt, highly practical advice: drop the polite pleasantries, feed it massive and complex multi-layered problems rather than a stream of micro-tasks, and stop wasting its compute on basic queries that don’t require a frontier model.
Part of the problem came from a leaderboard called Claudeonomics, which ranked employees by token use and gave out titles like Token Legend. Some employees reportedly kept Al agents running just to climb the rankings.
Now Meta is removing the leaderboard, adding a new usage dashboard called Al Gateway, and pushing engineers toward its own coding assistant, MetaCode.
Kang’s project won second place and a $175,000 prize at the 2026 Regeneron Science Talent Search. Experts still say the tool needs more research before it can be used in real clinics.
If validated, this could point toward faster and less invasive screening for neurodevelopmental conditions.
The eye may reveal the brain
A 17-year-old high school student built an AI tool that may help screen for autism and ADHD using retinal images. RetinaMind was created by Edward Kang from Bergen County Academies.
Agentic AI spurred a boom in mobile app releases, but there is no sign of these apps gaining traction, per FT.
Читать полностью…
A 5% stake in OpenAI would be worth about $42.6 billion, based on the company’s most recent valuation. The idea appears tied to a bigger question in Washington, how the public should benefit if AI companies keep growing in value.
This would be a major shift in how the U.S. government deals with private AI companies.
They are also helping improve the same AI tools that missed the mark. In other words, Ford is not replacing automation, it is adding human judgment where the systems still fall short.
The lesson is simple, in complex industries, experience still matters.
Ford is bringing humans back
Ford has hired, promoted or brought back around 350 experienced technical specialists over the past three years after its AI and automated quality systems did not meet expectations.
OpenAI just made AI cheaper.
The Information reports that OpenAI found a way to cut inference costs by more than half. Inference is the compute used every time a model generates an answer.
Anthropic says it has added new safeguards and will keep working with the U.S. government on testing and incident reporting.
Читать полностью…
The U.S. government has lifted export controls on Anthropic’s Claude Fable 5 and Mythos 5 models. Anthropic says Fable 5 will return globally across Claude, Claude Code, and cloud partners.
Читать полностью…
Candy AI is now on Telegram!
Candy AI
Candy.ai is the best AI girlfriend app, letting you create personalized virtual companions or connect instantly with our realistic AI characters in immersive, uncensored fantasy experiences - all within a safe and private space.
Ad. 18+
China’s AI is getting closer
A Chinese AI lab, Z.AI (Zhipu), released GLM 5.2, an open weight model that anyone can download and run. In Semgrep’s IDOR test, a benchmark for finding web access flaws, it scored 39% F1 and beat Claude Code, achieveing a Mythos Level result, while costing about $0.17 per vulnerability found.
The message is blunt, before AI can generate anything, the machines behind it need power, cooling, and resources.
The campaign uses fake Polaroid-style ads and handwritten notes to turn a technical issue into something people can instantly understand.