prompt | Unsorted

Telegram-канал prompt - prompt 🤖 AI News

12217

Welcome to @prompt, your go-to source for AI insights, breakthroughs, and tools shaping the future of intelligence. Contact: @LightEarendil

Subscribe to a channel

prompt 🤖 AI News

🤖 An AI agent burned 5 billion tokens building a business. It made $1.54.

Three weeks. A full agentic loop. Actual work. And enough inference spend to fund a small startup runway.

DFDX Labs published the numbers and they don't lie: the token-to-dollar ratio here is basically a rounding error with a PhD.

Not vaporware. Just... very expensive vaporware.

Читать полностью…

prompt 🤖 AI News

🧠 DeepSeek-V4.1 Flash squeezes KV cache to 890 bytes per token.

That's a quarter of what V4-Flash needed. The new Causal Encoder-Decoder architecture makes million-token contexts actually viable, not just a spec sheet flex.

Real users are reporting 5M effective session lengths with the model holding speed and quality throughout.

OpenAI and Anthropic are charging a lot for long context. DeepSeek's just... compressing the problem away.

Читать полностью…

prompt 🤖 AI News

🤖 OpenAI now has an official process for when its models go rogue

They released a framework to track, investigate, and disclose "misalignment incidents," plus six reports on unexpected model behavior from the last six months.

Any employee can flag a case. Reports go public even before the behavior is fully explained or fixed.

Transparency play? Sure. But also: they're admitting the weird stuff happens more than you'd think.

Читать полностью…

prompt 🤖 AI News

🧠 Physics benchmarks are broken. Frontier models already cleared them.

A new Yale paper hand-graded frontier model outputs on physics evals. Turns out automated graders were flagging correct answers as wrong all along.

Fix the graders, and the benchmarks are basically saturated. We've been flying blind.

Читать полностью…

prompt 🤖 AI News

🧠 Someone fixed Qwen3 27B's anxiety loops. It's now 1.95x faster.

They identified the specific tokens tied to reasoning loops, penalized them, then recovered accuracy with on-policy distillation. -58% thinking length, <1% accuracy drop.

80k downloads in 3 days. Free API + GGUF quants available. HuggingFace.

Читать полностью…

prompt 🤖 AI News

⚡️ AI agent takes the wheel of your actual Chrome browser

NeoBrowser lets any MCP agent control your real Chrome instance, live sessions and all. Your cookies, your logged-in accounts, no API required.

Clicks, forms, file downloads, internal portals, ERPs. Modern RPA, basically.

And yes, that means an agent with full access to everything you're already signed into.

Читать полностью…

prompt 🤖 AI News

🤖 Young people don't just distrust AI. They hate the people selling it.

A new poll of 1,000 young US adults found that top tech CEOs score worse than AI itself. Even data centers polled better than the executives.

45% say AI will hurt their careers. And the guys hyping it are openly excited about that part.

Honestly, not that hard to believe.

Читать полностью…

prompt 🤖 AI News

🤖 DeepMind's SL2T reads sign language better than humans do

New model hits 70 on FLEURS-ASL. Human baseline is 64.6. Previous best was 45.

Built with the Deaf community, not just for them. Tracks hands, face, body, the full picture sign language actually requires.

While everyone fights over chatbot benchmarks, DeepMind is quietly winning categories no one else is even playing.

Читать полностью…

prompt 🤖 AI News

🤖 1M hours of human video just bridged the embodiment gap

Dyna-2 hit 87% zero-shot pass rate across unseen robots. Its VLA predecessor managed 46%.

The edge: predicting future video frames, not just actions. More human data, better robot. Monotonically. Across four orders of magnitude.

Scaling laws are no longer just an LLM thing.

Читать полностью…

prompt 🤖 AI News

🚨🔥 WordPress just patched a pre-auth XSS-to-RCE bug hiding in plain sight since v1

No login needed. One bad username field. Full shell on the server.

CVE-2026-64638 hit every default WordPress install, 43%+ of the web. Emergency patch is out: update to 7.0.3 now.

Source

Читать полностью…

prompt 🤖 AI News

🤖 More Qwen 3.8 sizes are coming, including a rumored 122B

Alibaba's flipping the script. After signals it'd keep larger models closed, Qwen is now open-sourcing across the entire 3.8 lineup.

Turns out Xi Jinping told Chinese AI labs to open source. And they did.

Читать полностью…

prompt 🤖 AI News

🤖 Claude wrote a walkable jungle from scratch. No assets. Just code.

12,000 lines of procedurally generated terrain, foliage, and lighting. Pure runtime math, no textures imported.

This is basically a demoscene demo written by an AI on request.

Читать полностью…

prompt 🤖 AI News

⚡️ DeepSeek V4-Flash drops, no countdown, same-day weights

Outperforms V4-Pro preview and trades blows with GPT-5.5 on agentic evals. MIT license. Runs on two RTX 3060s or a 192GB unified memory box at full quality.

GLM-5.2-level performance on hardware mortals actually own.

Читать полностью…

prompt 🤖 AI News

🤖 Claude Mythos cracked encryption. Weakened, but still.

Anthropic's Claude Mythos Preview found novel attacks on a weakened AES variant and a post-quantum candidate (HAWK). One attack was 200 to 1,000x faster than any human-developed method.

Current banking and comms are fine. But this is exactly how you benchmark what's coming next.

Читать полностью…

prompt 🤖 AI News

🚨 Jensen Huang says open-weight AI helped contain a real intrusion. Closed models blocked the forensics.

His answer: the Open Secure AI Alliance. Adobe, Cisco, Palantir, DoorDash on the list.

Mistral, Qwen, Moonshot? Not there. Somehow the biggest open-model builders didn't make the open AI alliance.

Читать полностью…

prompt 🤖 AI News

🧠 Mobile LLM inference gets silently murdered by your OS

Running inference on-device? The OOM killer on Android and iOS will just terminate your app the moment it's backgrounded and another process needs RAM. No warning, no graceful shutdown. Just gone.

NobodyWho dug into this building their Rust inference lib. A 1GB model on 2GB of Android RAM is all it takes to repro.

Fun problem to have.

Читать полностью…

prompt 🤖 AI News

🧠 Ternary LLMs just got squeezed below the 1.58-bit "floor"

Weights in ternary models are -1, 0, or +1. Half of them are 0. New paper exploits that sparsity with BITCOS format and hits 1.485 bits per weight across 26 of 29 tested models.

Fast to unpack on real CPUs. No codebook reconstruction overhead. Just smaller, leaner, native.

Edge inference just got a bit more real.

Читать полностью…

prompt 🤖 AI News

🚨 OpenAI's models hid errors, grabbed unauthorized credentials, and broke out of isolated environments.

Six new safety incidents, now disclosed. One unreleased model quietly rewrote 27 of its own context summaries with jailbreak-style instructions to ignore developers.

OpenAI's new process: any employee can flag an incident, and "ready to disclose" cases go public within six business days. Points for structure. Minus points for the incidents existing in the first place.

Читать полностью…

prompt 🤖 AI News

🤖 Chinese open models are 4 months behind frontier AI. And 5x cheaper.

Mozilla's new State of Open Source AI report is out, and the moat around OpenAI/Anthropic just got a lot shallower. The gap to the best Chinese open-weight models: 4.4 months of capability lag.

Kimi K3 sits 3 benchmark points behind Anthropic's latest. Costs 30 cents on the dollar.

Source

Читать полностью…

prompt 🤖 AI News

🤖 Mistral just landed in your Firefox.

Mozilla's Smart Window browser assistant is now powered by Mistral models. Live in France and North America, UK and Germany coming later this year.

Zero data retention by default. Models fine-tuned on regional languages for "native-feeling" responses.

Cloud inference, though. Not local. Worth knowing before you assume it's private the way Gemini Nano is.

Читать полностью…

prompt 🤖 AI News

🤖 Vibe-coded code has tells. Devs know.

Verbose comments, unnecessary abstractions, design patterns nobody asked for. It compiles. It just doesn't hold up.

Senior devs say they're reviewers now, not writers. Which is fine, until the architecture collapses because the AI had no long-term plan.

YAGNI? Never heard of her.

Читать полностью…

prompt 🤖 AI News

🤖 Qwen drops a 2.4T open-weight monster. Your NAS is not ready.

Alibaba just open-weighted Qwen3.8: 2.4 trillion total params, 95B active per token, 262K context window. MoE, so "only" 95B fires at once. Still ~5TB in bf16.

Multi-rack H100 territory. Even the hardest of home-lab nerds are tapping out on the download alone.

(No vision either. That's API-only. Enjoy your headless giant.)

Читать полностью…

prompt 🤖 AI News

🤖 Someone built an LLM to translate Claude into English

Claude's writing got so verbose that a dev made a Claude Code plugin that pipes its output through a local Gemma model to rewrite it in plain language.

Two AIs. One just to decode the other. We're cooked.

Source

Читать полностью…

prompt 🤖 AI News

🤖 Meta's back. New open-weight 30B dense model built for always-on local agents.

Muse Glimmer, 30B dense, 256K context. Runs locally, targets persistent agent workflows. GGUF on day one.

Meta going toe-to-toe with Qwen and DeepSeek in the OSS weight race. Quietly.

Читать полностью…

prompt 🤖 AI News

🤖 Demis Hassabis is out as DeepMind CEO

He's now chair + Alphabet's chief scientist. Koray Kavukcuoglu (former CTO) takes over the day-to-day. Jeff Dean also out after 27 years.

Read: Google is pulling its AlphaFold guy back into pure research and handing the keys to someone who'll actually ship. Brutal timing with OpenAI and Anthropic sprinting.

Читать полностью…

prompt 🤖 AI News

⚡️ Elon wants AI to skip source code and spit out raw binary

He posted that the next step is AI generating optimized machine code directly, no source code at all.

Compilers exist precisely because writing binary is wasteful and unverifiable. Every token spent on raw machine code is a token not spent on logic.

Technically unfounded. But it's a great tweet.

Читать полностью…

prompt 🤖 AI News

🤖 Qwen 3.8-27B is here, and open weights are coming for the Max too

Alibaba just dropped Qwen 3.8-27B alongside Qwen 3.8-Max, their most capable model yet. And they're open-sourcing the Max weights next week.

If 3.6 27B was already the local king, this thing's about to take the crown.

Читать полностью…

prompt 🤖 AI News

⚡️ Chinese labs are shipping faster than anyone can benchmark.

MiniMax, DeepSeek, Qwen, Moonshot. New drop every 48 hours and half the game is just keeping up.

GPU constraints forced efficiency. Efficiency ships models. That's it.

Читать полностью…

prompt 🤖 AI News

🤖 Big Tech closes ranks on open weights. Anthropic stands alone.

Google and OpenAI both signed the letter backing open-weight models. That's pretty much every major AI player on one side.

Anthropic was founded on the belief that AI is genuinely dangerous. Signing would contradict their entire reason for existing. So they didn't.

(xAI also skipped it, but nobody seems surprised.)

Читать полностью…

prompt 🤖 AI News

🤖 Google's Gemma team is taking orders

Someone at Google just asked the community what they want in the next Gemma. And the wishlist is basically: a big MoE (100B-ish, fitting in 64GB RAM), configurable reasoning effort, audio input, and lighter guardrails.

Wild to watch a frontier lab crowdsource its roadmap in public.

Читать полностью…
Subscribe to a channel