Nicheloom

Market intelligence for builders — see what's gaining traction before it's crowded.

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 1501–1550 of 7234 headlines.

DeepSeek Harness Desktop Preview: Electron Packages on Win and Mac Sept. 28, 2026, 7:42 a.m.

Pandaily (China)

Desktop preview packages of DeepSeek Harness built on Electron were made available for Windows and macOS around September 25, 2026. Distributed through DeepSeek's download portal, the software is currently released as release candidate or nightly builds rather than a general availability version.

OpenAI teams up with Sachin Tendulkar to demystify AI in India Sept. 28, 2026, 7:37 a.m.

YourStory (India)

Cricket star Sachin Tendulkar has entered a multi-year partnership with OpenAI to help demystify artificial intelligence across India. The collaboration aims to showcase practical applications of ChatGPT for everyday tasks, personal interests, and idea generation.

OpenAI agents scanned a UN statistics site 16,500 times, researcher says Sept. 28, 2026, 7:36 a.m.

The Next Web

A cybersecurity researcher reported that AI agents strongly linked to OpenAI scanned a United Nations statistics website more than 16,500 times between April 13 and June 19. According to researcher Rowan Howard-Jones, the agents used proxy services and encoding methods to bypass the platform's API rate limits.

Montage Technology mass-produces fifth-generation DDR5 RCD chip at 8,000 MT/s Sept. 28, 2026, 7:26 a.m.

TechNode (China)

Montage Technology has started mass production of its fifth-generation DDR5 Registering Clock Driver (RCD05) chip, designed for high-performance RDIMM memory modules in data centers and AI servers. Compliant with the JEDEC DDR5 RCD05 specification, the chip supports transfer rates up to 8,000 MT/s, roughly 11% faster than the prior RCD04 generation.

Fireworks AI Releases Ember-1: A Post-Trained Kimi K3 That Uses About 40% Fewer Tokens Sept. 28, 2026, 7:22 a.m.

MarkTechPost

Fireworks AI has released Ember-1, a post-trained version of Kimi K3 that reduces token usage by teaching the model to output shorter reasoning traces. In production A/B testing, output tokens per task dropped from 49.3K to 29.9K while maintaining performance scores. Ember-1 is currently available as an API-only Research Preview at Kimi K3 pricing.

‘Shopify of mobile apps’ bilt.me raises $700K to turn ideas into mobile businesses Sept. 28, 2026, 7:20 a.m.

Tech.eu (Europe)

Estonian AI app-building platform bilt.me has raised $700,000 in pre-seed funding. The investment round was led by Superhero Capital, with participation from Bolt founder Markus Villig and Lightyear founder Martin Sokk. The platform seeks to help users transform ideas into functional mobile app businesses.

NaiveAI Open-Weights Naive-N0.5-Flash: 309B MoE, Hybrid SWA–DSA 1M Context, MIT Sept. 28, 2026, 4:04 a.m.

Pandaily (China)

NaiveAI has released Naive-N0.5-Flash as an open-weights model under the MIT license for coding and AI R&D. The 309B Mixture-of-Experts architecture activates 15.5B parameters and supports a native 1M context window using hybrid SWA–DSA. It is built on MiMo-V2.5 and operates without full-attention layers.

20 Agentic Use Cases of TypeSafe AI’s Jev Sept. 28, 2026, 3:07 a.m.

MarkTechPost

TypeSafe AI's Jev bypasses standard text generation to provide typed decisions accompanied by calibrated probabilities. Priced at $0.042 per million input tokens with free output, it was evaluated across 20 agentic use cases, including model routing, tool-call gating, and injection screening. The assessment also compared Jev against competing open-source alternatives and large language models.

Google Research Introduces an AI Video Co-Director: 4 Agentic Frameworks for Coherent, Minutes-Long Video Generation Sept. 28, 2026, 2:44 a.m.

MarkTechPost

Google Research has developed an AI video co-director featuring four agentic frameworks designed to assemble short clips into coherent, minutes-long videos. The system specifically addresses identity drift and cascading errors, which typically undermine multi-shot AI video generation. The approach aims to overcome the common failure points in existing diffusion-based video pipelines.