AI news for builders and product teamsUpdated Nov 17, 2026, 14:00 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
ArtCraft, founded by Brandon Thomas, has released 17 open source clones of popular creative tools, starting with a Rust-built Photoshop alternative called PhotoCraft, and says its apps are downloaded about 1.7 million times a day. At the AI Engineer Code Summit he described using large AI prompts to build the apps and outlined an ambition to rebuild commercial software as open source.

Simon Willison released llm-gemini 0.37, adding support for Google's new Gemini 4 Hydrogen, the cheapest Gemini 4 model at $1.50/$7.50 per million input/output tokens, though only 25% below Carbon and Argon until their prices double on December 1, 2026.

GitHub published a practical guide to agent skills — versioned directories built around a SKILL.md file that capture repeatable task instructions — based on a GitHub Universe 2026 session with Shishir Tewari, Senior Manager of Data Engineering at Procore.

Democrats on the House Ways and Means and Senate Finance committees opened an inquiry into how the Trump administration handled Nvidia's H200 chip sales to China, sending a letter to Commerce Secretary Howard Lutnick seeking preservation of records. The move came the same day Nvidia ousted CEO Jensen Huang, who agreed to testify before a separate Republican-led group of committees.

The FAA has expanded its $875 million SMART AI system, which advises air traffic controllers on routing and scheduling, from the Washington, DC region to the airspace of all 22 US air traffic control centers. The agency says the tool rerouted 180 flights in its initial deployment, while the controllers' union says it was not involved in designing or testing it.

Meta's consumer AI agent Muse can now place outbound calls to U.S. businesses on users' behalf, but the feature is limited to paying subscribers after an earlier test with select early adopters. Muse plans start at $20 per month.

Amazon announced that SGLang-Neuron, released by RadixArk with Annapurna Labs, is publicly available through an AWS Deep Learning Container, letting teams serve models and run reinforcement learning rollouts on AWS Trainium through the standard SGLang interface.

Natera built SWARM, an agentic AI system on Amazon Bedrock AgentCore, that cuts average first response time on lab equipment support tickets from 17.4 hours to roughly 8 minutes (about 130x faster) and gives 96.2 percent of tickets an automated first response.

Microsoft Research's newsletter roundup reports that a new study of 13.5 million coding sessions found fewer than 8% needed context compaction but those sessions consumed 44% of AI processing costs, and introduces CompactionBench plus APEC, which cut costs by up to 25.4%.

Nebius disclosed $40 billion in planned spending for 2027 on its third-quarter earnings call, to be funded largely by customer prepayments and loans backed by customer contracts. Its Microsoft contract, worth up to $19.4 billion, is already slipping, with only just over half of that capacity now due by year-end and the rest in Q2 2027.

An AWS post details how four request-level Amazon Bedrock settings — prompt cache TTL, Flex service tier, adaptive thinking effort and maxTokens — change token counts and rates without touching model or prompt. In its Opus 5 agent test, low effort used 936 median output tokens versus 2,521 at the default high.

Nvidia representatives privately told Representative Bill Foster that the company could technically implement the geotracking requirements in the Chip Security Act, according to Foster, even though the Semiconductor Industry Association, which Nvidia belongs to, called such mandates "new, untested, and potentially infeasible."

Amplitude uses OpenAI models to power Wave, a product agent that spots what to build, helps teams ship changes via coding agents, and measures results, cutting improvement cycles to one or two days.

Amplitude's product agent Wave, built primarily on OpenAI models, is described as identifying what to build, helping teams ship changes via coding agents, and measuring impact — reportedly cutting shipping time to one or two days.

Ars Technica is hosting a live discussion on November 17 at 1 pm ET about drone and missile strikes on data centers in Eastern Europe and the Middle East this year, which disrupted cloud services for millions and permanently deleted some AWS customer data.

Google released Gemini 4 Hydrogen, its replacement for Gemini 3.8 Flash, which scores 51 on the Artificial Analysis Intelligence Index at the default thinking level — edging out the larger Gemini 4 Argon at 50 and reaching 54 at a new xhigh tier. It costs $1.50 per million input tokens and $7.50 per million output tokens.

Microsoft says a latent defect from a July Microsoft 365 networking gateway update, combined with a Teams failover and roughly 14% of gateway capacity already offline, caused a cascading outage that left over 80% of servers unresponsive at its peak.

AWS published a no-code pattern for attributing Amazon Bedrock costs to individual IAM principals using automatic identity tagging, cost allocation tags, CUR 2.0 or Cost Explorer, and AWS Budgets actions that can attach a Deny policy when a threshold is crossed.

Google is rolling out Gemini 4 Hydrogen to Search AI Mode worldwide, claiming it is the biggest Search model improvement yet and that it handles multi-part or typo-heavy questions faster than the previous fast model, Gemini 3.8 Flash.

IBM's Granite 4.1 3B and 8B models now have experimental LLMON previews that use reserved special-token delimiters to separate instructions from data, cutting indirect prompt-injection success rates close to zero on several benchmarks while matching standard Granite 4.1 Instruct performance.