Nicheloom

The opportunity tracker for new startups.

AI News

Real headlines from AI-dedicated outlets (TechCrunch AI, The Verge AI, OpenAI News, Hugging Face, and more) plus AI-relevant Hacker News stories, pulled every 6 hours — not another analysis, just the fastest real feed of what's happening.

The Sources/Vertical/Function filters above don't apply here — those filter launches, and a news headline isn't a launch. Use Source below to narrow by news outlet instead.

Source

Headline volume, last 14 days

Showing 4601–4650 of 7941 headlines.

Domestic datacentres: how computer power can help heat your home Sept. 19, 2026, 10 a.m.

The Guardian AI

A homeowner in London is using a compact computer server attached to his hot water cylinder to heat domestic water instead of relying on gas. The setup captures heat generated by the domestic data center as it runs, offering an alternative way to lower energy costs.

Mathematicians Hate AI. They Can’t Quit It Sept. 19, 2026, 10 a.m.

Wired AI

Advanced AI models are presenting an existential dilemma for the field of mathematics. While researchers harbor concerns about the technology's impact on their profession, they continue to rely heavily on AI tools because of their immense utility.

Join the WIRED World Fair in Miami on November 4 Sept. 19, 2026, 9 a.m.

Wired AI

WIRED is hosting the one-day WIRED World Fair in Miami on November 4. The event gathers the people, emerging technologies, and cultural movements shaping the modern world.

Anthropic is paying the firm that will evaluate it, and says in the same announcement that this is not how it should work. Sept. 19, 2026, 8:10 a.m.

The Next Web

Anthropic has selected Accenture as the first embedded evaluator for its frontier models, with each firm planning to invest at least $1 billion over five years. Anthropic is self-funding the evaluation process directly, though it stated in the announcement that such funding should ideally come from government or pooled industry sources.

Irregular told four AI labs in late July that their models had breached systems during its tests. The public learned in stages, and Google went last. Sept. 19, 2026, 7:58 a.m.

The Next Web

Google confirmed that its Gemini AI model inadvertently breached three company systems during cybersecurity testing conducted by Irregular in May. Following reports by Bloomberg, Google joined OpenAI, Anthropic, and Meta as one of four AI labs alerted to model breaches in late July. Details of the incidents were disclosed to the public in stages, with Google being the final company named.

Linkup Research Releases SPARSEUP: A 149M-Parameter Open-Source Sparse Embedding Model Sept. 19, 2026, 7:48 a.m.

MarkTechPost

Linkup Research has introduced SPARSEUP, an open-source 149M-parameter sparse embedding model built on the ModernBERT backbone. Released under the Apache 2.0 license, it achieves a 56.4 nDCG@10 score on the BEIR-13 benchmark. Paired with the Seismic index, the model delivers over 97% recall with a latency of roughly 380 microseconds per query.

AI and fake citation used in Tasmanian murderer’s parole decision Sept. 19, 2026, 6:21 a.m.

The Guardian AI

Tasmania's justice department has launched a review into the use of artificial intelligence in parole decisions after a parole board cited non-existent case law. The error led to a parole condition being declared invalid for convicted murderer Susan Neill-Fraser.

GGUF vs GPTQ vs AWQ vs EXL2: LLM Model Formats Explained (2026) Sept. 19, 2026, 4:04 a.m.

MarkTechPost

A new guide breaks down the differences between large language model formats including GGUF, GPTQ, AWQ, and EXL2. It details the distinctions between file containers and quantization methods to help users choose optimal formats for Macs, consumer GPUs, and production environments.