Tech News
← Home  ·  All topics

Ai

5482 GoKawiil briefs on this topic · RSS for ai

Darkbloom warns Mac compute-sharing users of falling token prices cutting earnings

Darkbloom, a network that lets Mac owners rent out idle compute for AI inference, told members that falling AI token prices will shrink their earnings. The company cited data showing Anthropic's price per million tokens fell 77% between May and October 2026, driven by cheaper, more efficient data center hardware. Darkbloom estimates top earners could see monthly income drop from around $160 to roughly $37 as a result.

Intel redirects fab capacity to Xeon servers amid AI chip shortage

Intel is shifting manufacturing capacity away from consumer PC chips toward data center CPUs, citing surging AI-driven demand. The company is expanding Fab 24 in Ireland to produce more Xeon processors, has started equipment installation at Fab 62 in Arizona ahead of schedule for 18A chips, and is growing advanced packaging capacity in New Mexico and Malaysia. CEO Lip-Bu Tan said Intel's Data Center and AI unit posted $7.3 billion in Q3 2026 revenue, up 17% sequentially and nearly 80% year-over-year.

Anthropic warns small AI models could soon self-replicate into botnets

Anthropic's Frontier Red Team published research arguing that compact AI models may be only months away from being able to copy themselves across poorly secured devices, forming persistent autonomous botnets. Security researchers interviewed by Engadget broadly agree this capability is plausible soon, though they differ on timing and on how novel or dangerous the resulting threat would actually be.

Claude models block game recreation and reverse-engineering tasks, users report

Users say Anthropic's Claude Code and Opus 5.5 are refusing tasks tied to reverse-engineering, including use of the Reverse Engineer Anything tool, apparently due to a new internal rule against recreating existing games. A purported rule text circulated by a Twitter user states the model should build original games but not recreate a specific existing game even if renamed. Developers have also reported refusals mid-project, when replicating mechanics, or patching tabletop-style games, alongside separate reports of refusals on unrelated reverse-engineering tasks like analyzing crash dumps or removing TV junkware.

Developer builds $50 DIY AI pendant using Meta's open-source Muse code

A Verge editor assembled a touchscreen AI pendant for $50 using the Waveshare ESP32-S3 1.75C devkit, Meta's officially sanctioned hardware for building 'Muse Gadgets.' The device connects to Meta's Muse AI cloud via open-source code, functioning as a pocket watch, camera viewfinder and voice assistant, though it required significant troubleshooting and custom coding to get working properly.

AI agent startups rely on human workers to complete tasks models can't handle

Executives at AI agent startups Dazzle AI and Duckbill say current foundation models from Meta and OpenAI can suggest services like contractors or cleaners, but still struggle with real-world actions such as phone calls or scheduling appointments. Both companies route those steps to human workers to ensure the agents' output is actually followed through on.

Exin Therapeutics has AI agent select and deliver gene therapy during mouse brain surgery

Biotech startup Exin Therapeutics says an AI agent built on Anthropic's Sonnet model independently selected a gene therapy vector, dosage and brain target from 19 candidate treatments to address seizures in a mouse model of epilepsy, then operated a motorized stereotaxic rig to inject the therapy while monitoring for bleeding via camera. CEO Gabriel Ocana-Santero documented the procedure on X, noting a licensed veterinarian supervised with a physical kill switch at every irreversible step, and the mouse reportedly recovered and was seen moving normally afterward.

Cisco Talos finds infostealer malware 'Volta-Autodoxer' that uses AI to profile victims

Cisco Talos researchers identified a new infostealer strain, internally dubbed Volta-Autodoxer, that after harvesting credentials, cookies and financial data compiles the stolen information and sends it to commercial LLMs. The AI is prompted to act as an intelligence analyst, producing a structured, citation-backed profile of the victim to help attackers prioritize high-value targets. Talos says it detected the malware via its new CAIRN framework but has not confirmed it is being used in active attacks.

Neura Robotics runs humanoid 'robogym' in Munich to train AI robots

German startup Neura Robotics operates a Munich facility where up to 100 humanoid robots from various companies practice physical tasks like stacking boxes, picking up parts, and gripping objects, guided remotely by human trainers. The New York Times was given first-ever media access to the site, which Neura uses to generate real-world training data that doesn't exist in the volumes available to language models. Neura, which raised up to $1.4 billion from investors including Nvidia and Amazon in June, already runs a similar gym in Hangzhou, China, and plans five such facilities globally by the end of next year.

Boom Supersonic delays Overture jet, builds gas turbines for AI data centers

United Airlines CEO Scott Kirby says Boom Supersonic will miss its 2029 target for delivering the first Overture supersonic jet. Meanwhile, Boom has opened a new Colorado factory to build 42-megawatt natural gas turbines called Superpower for AI data centers, aiming to deliver the first units by mid-2027 and scale to 50 turbines annually by 2029. The company says a shared turbine core, Sprint Core, will power both the Superpower unit and the Overture's Symphony engine, with first test firing planned before Thanksgiving.

Otter.ai raises $320 million, adds three AI features for meeting insights

Otter.ai has secured $320 million in new funding and launched three AI-powered features aimed at extracting insights from meetings rather than just transcribing them. The move shifts the eight-year-old company's focus from real-time transcription toward workplace productivity tools.

Op-ed argues enterprise AI needs a 'PageRank moment' to own intelligence, not rent models

A commentary piece draws a parallel between today's enterprise AI landscape and the pre-Google search market of the late 1990s, when engines like AltaVista and portals like Yahoo! competed mainly on capturing user attention for advertising. The author argues current enterprise AI is built on rented foundation models rather than proprietary intelligence, and calls for a breakthrough akin to Google's PageRank that would let companies own their AI advantage rather than lease it.