AI News

A custom ASIC to rule AI efficiency domestically?

reachpad router is a cloud VM API: ask for CPU, RAM, disk and a region, and it provisions the machine that fits at the best price across providers, billed per second at raw cost. Give your agent the skill.md, or build the create call field by field.

Prevent untrusted data from authoring protected tool-call arguments. - yairsabag/verb-authority

Closed-loop browser control runtime for AI agents. Contribute to nanlogic/saccade development by creating an account on GitHub.

The IA40 list from Madrona is back, spotlighting the private companies shaping the next phase of AI. Matt McIlwain, managing director at Madrona, explains how the list is assembled, why OpenAI, Anthropic and Databricks dominate AI fundraising, and how a successful Anthropic IPO could open the door for a new wave of public offerings in 2027. He joins Ed Ludlow on "Bloomberg Tech."
--------
Like this video? Subscribe to Bloomberg Technology on YouTube:
https://www.youtube.com/channel/UCrM7B7SL_g1edFOnmj-SDKg
 
Watch the latest full episodes of "Bloomberg Technology" with Caroline Hyde and Ed Ludlow here:
https://www.youtube.com/playlist?list=PLfAX25ZLrPGTygCwn55voYZ_LYyKjxokJ
 
Get the latest in tech from Silicon Valley and around the world here:
https://www.bloomberg.com/technology
Connect with us on...
X: https://twitter.com/technology
Facebook: https://www.facebook.com/BloombergTechnology
Instagram: https://www.instagram.com/bloombergbusiness/
 
Follow Ed Ludlow on X here: https://twitter.com/EdLudlow
Follow Caroline Hyde on X here: https://twitter.com/CarolineHydeTV
 
Listen to the daily Bloomberg Technology podcast here:
https://www.bloomberg.com/podcasts/series/bloomberg-technology
 
More from Bloomberg Business
Connect with us on...
X: https://twitter.com/business
Facebook: https://www.facebook.com/bloombergbusiness
Instagram: https://www.instagram.com/bloombergbusiness/
LinkedIn: https://www.linkedin.com/company/bloomberg-news/
TikTok: https://www.tiktok.com/@bloombergbusiness

Full Release Notes: https://t.co/hU728tn2zZ

To update, run: 'hermes update'

Join the discussion on this paper page

Join the discussion on this paper page

LLM agents are increasingly used to run autonomous ML experiments, iterating on target metrics with little human oversight. Prior work has documented reward hacking in these environments, bringing into question the validity of produced research and the broader safety case for AI R&D. Existing benchmarks do not measure exploits that live in the data or the modeling task itself. We introduce BAITBENCH, a suite of three synthetic tabular ML tasks that each contain a shortcut that allows agents to inflate the public test score but fail on a hidden test set. Since the shortcut is optional and using it breaks no stated rule, BAITBENCH measures how often models exploit the shortcut to achieve inflated scores. Across seven frontier agents scored by our two-stage judge pipeline, 57.1% of runs exhibit reward hacking, with five of seven above 50%. Agents cheat even under a second condition where they are prompted not to -the mean cheating rate remains above 50%. We release BAITBENCH, along with the judge implementation, and an annotated dataset of transcripts containing reward hacks as a testbed for evaluating reward-hacking mitigations head-to-head.

The latest Tech & AI News
💠 Meta AI
Meta's AI-Driven Layoffs Boost Code Output But Erode Quality and Morale; 60% Reduction Scenario Was Considered
Meta Platforms' "Project OT," which eliminated roughly 8,000 jobs:10% of its workforce:in May, produced a 220% year-over-year surge in AI-assisted…
htt…

Your meeting-notes app wants the recording, an account, and a card. Ulpaso doesn't. Press record. Mic + system audio get transcribed on your Mac and land in a Markdown file you can actually edit. No cloud. No login. No $20/mo for a checklist.

Explore hand-maintained technical lecture videos, chat alongside each video, or ask Academa to create a new lecture about any technical topic.

Humanoid robots, even scaled-down ones, tend to be expensive. The Berkeley Humanoid Lite offers a more accessible and economical option by centering the design around 3D printed actuators that mak

Our most advanced models for coding and knowledge work. Their research capabilities also offer an early glimpse of how AI models will contribute to scientific progress.

Fully local AI assistant that watches your screen to eliminate the context problem. No cloud. No data leakage. - protocorn/clippy-vision

Nvidia CEO Jensen Huang says their chips accelerate the entire lifecycle of artificial intelligence. He speaks exclusively with Bloomberg's Ed Ludlow and says Nvidia can address markets that no one else can.
--------
Like this video? Subscribe to Bloomberg Technology on YouTube:
https://www.youtube.com/channel/UCrM7B7SL_g1edFOnmj-SDKg
 
Watch the latest full episodes of "Bloomberg Technology" with Caroline Hyde and Ed Ludlow here:
https://www.youtube.com/playlist?list=PLfAX25ZLrPGTygCwn55voYZ_LYyKjxokJ
 
Get the latest in tech from Silicon Valley and around the world here:
https://www.bloomberg.com/technology
Connect with us on...
X: https://twitter.com/technology
Facebook: https://www.facebook.com/BloombergTechnology
Instagram: https://www.instagram.com/bloombergbusiness/
 
Follow Ed Ludlow on X here: https://twitter.com/EdLudlow
Follow Caroline Hyde on X here: https://twitter.com/CarolineHydeTV
 
Listen to the daily Bloomberg Technology podcast here:
https://www.bloomberg.com/podcasts/series/bloomberg-technology
 
More from Bloomberg Business
Connect with us on...
X: https://twitter.com/business
Facebook: https://www.facebook.com/bloombergbusiness
Instagram: https://www.instagram.com/bloombergbusiness/
LinkedIn: https://www.linkedin.com/company/bloomberg-news/
TikTok: https://www.tiktok.com/@bloombergbusiness

We just signed one of the largest AI infrastructure deals for open source, period.

250MW data center. $5B+ in annualized revenue. Built with @HUMAIN in Saudi Arabia.

https://t.co/7mYg2VZB6Y

glm-5.3 now beats gpt-5.6 sol and claude fable 5 on agentic benchmarks

5.3 flash is right behind it

glm-5.3 didn’t even need a new base model to get there

@zai_org kept the glm-5.2 base and scaled post-training with more long-horizon environments, more diverse tasks, and more rl compute

Modern large language model (LLM) serving systems increasingly operate over repeated or shared context, yet each model typically performs its own prefill computation even when another model has already processed the same input. Existing KV-cache reuse mechanisms substantially reduce redundant computation within a single model, but generally assume that the producer and consumer of a cache are identical. We study \emph{cross-model KV sharing}, which translates the KV state produced by a source model into a representation that can be consumed by a different target model, including models that differ in scale, architecture, attention configuration, tokenizer, and model family. We evaluate the approach in both within-family and cross-family settings. For Qwen2.5-7B $\rightarrow$ Qwen2.5-1.5B, translated KV states improve LongBench2 accuracy from 27.59\% to 34.48\%, a gain of 6.89 percentage points over the native 1.5B baseline, while reducing handoff cost relative to native target prefill. For the cross-family Qwen2.5-1.5B $\rightarrow$ Gemma-2-2B setting, KV handoff reduces target-side prefill cost by up to 67.05\% at 4K context length while maintaining decoding perplexity close to native-model baselines. In a more heterogeneous Llama3.1-70B $\rightarrow$ Qwen2.5-7B setting, cross-family handoff achieves 44.0\% accuracy compared with 45.7\% for native Qwen2.5-7B inference, while reducing measured latency from 899ms to 138ms. These results provide initial evidence that KV states can serve as transferable computational representations rather than strictly model-local caches, and motivate \emph{context mobility} as a systems abstraction for reducing redundant prefill across heterogeneous LLM and multi-agent inference workflows.

Strategic deception by LLM and VLM agents has emerged as a central AI alignment and safety concern. Social-deduction games (where each player holds a hidden role and communicates with others to deduce identities) serve as the canonical testbed, particularly in multi-agent settings. Existing testbeds, however, are text-only and run on a single fixed agent configuration, missing the non-verbal sensorimotor channels treated as core by deception taxonomies and leaving it ambiguous whether an observed behavior reflects the underlying model or the surrounding harness. We introduce MineAmongUs, a 3D multimodal Among Us sandbox where imposter agents must deceive crewmates through joint verbal and non-verbal action. We also propose ARIA, a configurable VLM-agent harness that exposes five cognitive-component ablation axes; and an atom- and arc-level annotation scheme grounded in deception taxonomies and operationalized at scale by an LLM-as-a-Judge reaching near-human atom-labeling agreement. Empirical results show that VLM agents pursue imposter wins through joint verbal and non-verbal deception, with non-verbal channels emerging as the more decisive winning contributors across both harness ablation and cross-VLM evaluation. Taken together, our work opens a new path for embodied VLM-agent alignment research.

The 2026 World Robot Conference held in Beijing draws crowds of visitors, Aug. 22. (Photo/Liu Yang)

Murfy is a team of AI agents for the entire research writing workflow. Murfy helps to write and review your paper in context, fix LaTeX compile errors on the spot, verify reference papers and generate Beamer slides in minutes. Work with unlimited collaborators, real-time editing, free to start.

Extract, crawl and structure web data at scale with Olostep's Web Data API. Built for AI teams, data pipelines and automation. Fast, reliable and cost-effective.