AI news for builders and product teamsUpdated Nov 18, 2026, 10:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
IEEE Senior Member Balaji Ingole is a project manager at B2B e-commerce company Amla in Milwaukee, where he develops AI tools, including chatbots that help manufacturers set up and manage large product catalogs. He also conducts independent research in AI-enabled health care applications.

Amazon released SOP-Bench, an openly available benchmark that measures how well AI agents execute real standard operating procedures authored by domain experts across 12 business areas, with more than 2,000 tasks, functioning tools, and ground-truth answers. Testing two baseline agents across 11 frontier models showed that newer models sometimes scored lower, extra tools nearly halved success on one procedure, and no single model-agent pairing won everywhere.

IEEE Spectrum is promoting a free webinar on using agentic AI for semiconductor root cause analysis, featuring a live demonstration with Spotfire Industry Pro. The session covers cross-domain analytics, automated visualization, and scaling high-performance analytics across large fab datasets.

Ben's Bites describes how the author set up a fresh personal agent using a folder of instruction and memory files, then revised the plan after testing suggestions from a coding-focused agent. Around 700 readers said they use or want a personal agent, and roughly 50% of poll respondents asked how it works in Claude.

Hugging Face research introduces three tests to quantify benchmark optimization in speech recognition, finding that several top-scoring open-source ASR models reproduced benchmark reference transcripts even when the audio contradicted them, words were silenced, or multiple written forms were equally supported.

Hugging Face detailed how Papers with Code uses Hugging Face Jobs, Storage Buckets, and Inference Endpoints to build a hybrid search system combining PostgreSQL full-text search with pgvector semantic search, covering more than 110,000 papers from arXiv and Daily Papers.

Together AI compared GLM-5.3 and GPT-5.6 Sol across 904 DeepSWE rollouts on 113 tasks. Sol led pass@1 at 72.7% versus 69.0%, while GLM-5.3 won pass@4 and cost about 2.1x less per rollout; a GLM-first cascade with Sol escalation solved 85.9% of tasks at $6.61 each.

MIT researchers developed a computational approach to predict promising catalysts for electrochemical ammonia production, published Aug. 11 in EES Catalysis. The work is purely theoretical so far; the materials still need to be made and tested.

Liquid AI released DSpark draft model checkpoints for three LFM2.5 models, adding speculative decoding that it reports delivers up to 3.18x throughput improvement on GPU and up to 2.87x on-device without changing output quality.

Ben's Bites reports a poll with 996 total votes on personal agent use, and describes personal agents as a folder of files and tools connected to any agentic tool. Headlines include Meta AI's Mac app, Berd by Block, OpenAI pausing model improvements for two weeks, and Ramp launching router.com.

A Hugging Face blog post reports that ALTK-Evolve, which distills guidelines from an agent's own past trajectories and injects them at inference time without weight updates, improves task completion only when the amount of memory is calibrated to the model. Across eight models, strong models benefited from the full guideline set while weaker models did best with curated retrieval.

MIT CSAIL researchers describe a phenomenon they call attribution decay, in which larger training sets weaken any individual example's influence on a diffusion model's output. Their exact deletion method, published in Nature Communications, found the counterfactual radius shrinks as datasets grow.

Ben's Bites is asking readers whether they use agent products as a personal agent, with answers to be revealed in Thursday's post. The newsletter's headlines also cover Grok Bot, new Codex and Claude Code features, Gemini 3.7 Flash, and Stripe's agreement to acquire OpenRouter for more than $7B.

Sentence Transformers v6.0 adds a fourth model type, MultiVectorEncoder, for ColBERT-style late interaction retrieval, per a Hugging Face blog post. The post says PyLate and Stanford-NLP ColBERT checkpoints load directly, colpali-engine visual document retrieval models can be used after per-repository configuration, and it describes MaxSim scoring plus index-size tradeoffs.

MIT professor Eugene Fitzgerald discusses his new book "The Invisible Engine: Why Innovation Evades Control," which examines how innovation actually happens and how society should invest in it. The interview covers his strained silicon co-invention at Bell Labs, three categories of research investment, and his argument that innovation is a process rather than an idea.

Dharma AI benchmarked a constraint-aware GPU allocator against a FIFO scheduler across seven scenarios, reporting GPU utilization gains of up to 33 percentage points and priority-weighted output increases of up to 105% on identical hardware and workloads.

Nathan Lambert argues Nvidia is funding open-source model development, reportedly spending $26 billion, so that many companies can build and run their own models rather than buying tokens from Anthropic or OpenAI. He outlines two possible futures for open models and says a long-tail specialization outcome is his most likely.

AMD says it surpassed its AI productivity goal, reaching a 30 percent overall boost, and reports that AI agents now resolve more than 75 percent of issues in its Radeon Software eXperience component, up from 6 percent in October 2025.

A team at Axiom Math says its AxiomProver system automatically verified the proof of the "246 theorem" about prime numbers, a first for the company's AI system. The work formalizes the Polymath8b result that there are infinitely many primes differing by 246.

Together AI describes running A/B experiments at the endpoint level, splitting live traffic among one control and up to 20 variants with fixed percentages. It covers ramping via member updates, measuring platform and product metrics per deployment, and ending tests by promoting a rollout or deleting the experiment.