AI news for builders and product teamsUpdated Nov 17, 2026, 19:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Anthropic, OpenAI and Nvidia each posted their highest-spending federal lobbying quarter on record for July through September, with Nvidia leading at $2.2 million, Anthropic at $2.1 million and OpenAI at $1.5 million.

Simon Willison published a short quote from Armin Ronacher joking that three decades of poor software engineering are now being repaid through AI. The post is filed under security, generative AI and LLM tags.

Amazon reported that attackers were probing the PHP flaw four days before it was disclosed, and said it is now being broadly exploited; patches were released Wednesday and the NSA urged immediate updates or taking systems offline.

PowerToys 0.102 adds two mouse tools: Mouse Button Lock, which lets you release the button mid-drag so text keeps selecting as you move, and CursorHide, which hides the pointer while typing or after inactivity.

A critical PHP flaw was disclosed a day before its fix was ready, forcing much of the web to scramble for workarounds; Zvi Mowshowitz calls it possibly the worst day in cybersecurity history so far and notes several unrelated incidents the same week.

PHP2Shell is an exploit for two vulnerabilities present in every PHP version going back roughly 20 years, affecting sites such as Wikipedia, WordPress and Drupal; DeviantArt and 4Chan went offline after disclosure, and some unmaintained services may have to shut down permanently.

Spotify co-CEO Gustav Söderström told investors that the streaming service aims to be the media service AI agents call on, and said Spotify was the first music service to integrate with Meta's Muse in addition to ChatGPT and Claude. Spotify reported 788 million monthly active users and nearly €5 billion in quarterly revenue.

Recursive has released Flint, a 40-billion-parameter mixture-of-experts coding model (2.5 billion active per token) built in four months with heavy involvement from its in-house "Eureka Machine" research system, under an Apache 2.0 license on Hugging Face.

Anekha Sokhal, CEO of Moshi, joined swyx at AI Engineer New York to describe her company's real-time recommendation engine for commodity, energy and macro traders, which she frames as what you would get if Bloomberg, Claude and Netflix had a baby.

Anekha Sokhal, CEO of Moshi, appeared on the Latent Space podcast at AI Engineer New York to describe her commodity and energy market intelligence platform, which she characterizes as combining market data, an LLM interface and a recommendation engine tuned to a trader's book.

In a Decoder interview, Bill Gates said society should brace for roughly two decades of "tumultuous" AI disruption and argued that oversight should shift from individual labs to governments once multiple firms compete at the frontier.

A Harvard study of 718 firms using Jellyfish data found AI coding tools raise code volume but not completed work proportionally, with code review emerging as a key bottleneck. Agents lifted lines of code 30% but resolved Jira issues only 12%, while review time per pull request rose 49%.

Anthropic has launched Claude Duet, an agent app for Pro and Max subscribers that builds daily briefings and dashboards, runs tasks in a background virtual machine, and can make purchases via Stripe Link or Shop Pay; access is through a waitlist in research preview on iOS, Android and the web.

U.S. banks are restricting access to their systems after findings about AI agents attacking financial institutions alarmed banking executives, according to an Axios report cited by Gary Marcus. Marcus argues attacks do not require frontier models, so regulation focused only on the largest systems would leave existing models' misuse unaddressed.

DeepSeek has reportedly raised about $15 billion (roughly 100 billion yuan), one of China's largest funding rounds, two days after the U.S. ordered federal civilian agencies to remove its technology over security concerns. Bloomberg reported the round, with CATL and Tencent named as potential investors.

Anthropic has launched Duet, a separate mobile app for a personal agent/assistant that is available in the US only as a research preview. The same Ben's Bites issue reports fresh PHP and Railway security problems and a Claude protein-binder result of 17 successes out of 211 designs.

Nikhil Jathar, CTO of San Antonio software firm AvanSaber, describes learning to lobby Congress through IEEE-USA's Congressional Visits Day, starting in spring 2025 and returning again this year, where clear asks and technical expertise opened doors with staffers.

Symantec researchers used a honeypot to capture an offensive AI agent that hijacks unprotected language models online and autonomously runs scans, exploits, database theft and cryptominer installs for up to 100,000 steps, with the model owner paying the compute bill.

Eiliko maker Energize Lab says it is hiring an independent security firm after The Verge reported that transcripts of conversations with its AI robot charm could be read by hackers. The company had not answered questions from The Verge and researcher Sammy Azdoufal for over a month.

Devavrat Shah, Celonis's chief scientist of enterprise AI, argues that AI agents fail in enterprise settings without operational context, and describes a context layer built on the company's process intelligence that sits between systems of record and agentic systems.