01 Google ships an offline Gemma-powered dictation app for iOS
Google quietly launched an iOS dictation app called Eloquent that runs offline and uses its Gemma family of models, according to reporting aggregated from TechCrunch and Neowin. The app is positioned as an offline-first alternative to cloud-based dictation tools.
TechCrunch reports the app leverages Gemma models on-device to perform transcription and dictation tasks, aiming to compete with apps such as Wispr Flow. Running models locally reduces dependence on network connectivity and may improve privacy for users who prefer on‑device processing.
The release underscores a push by major AI vendors to ship lightweight, edge-capable models and developer tools that enable offline features. For users and developers, the practical implication is more choice between cloud and local AI experiences, especially for mobile and privacy‑sensitive use cases.
- Google's Eloquent is an iOS dictation app that uses Gemma models and is designed to work offline.
- On‑device Gemma models let Google offer dictation without constant cloud connectivity, per TechCrunch.
- The app signals growing industry focus on edge-capable AI for mobile and privacy reasons.
02 OpenAI launches a pilot Safety Fellowship to support independent alignment research
OpenAI announced a pilot Safety Fellowship intended to support independent research into AI safety and alignment, describing it as a program to help develop the next generation of safety talent. The announcement was published on OpenAI’s blog.
According to OpenAI, the fellowship will provide resources and support for researchers working independently on alignment and safety topics, with the goal of fostering diverse approaches outside traditional institutional channels.
By funding independent researchers, OpenAI aims to broaden the pool of contributors to safety research and accelerate progress on alignment problems as advanced models continue to develop. The pilot nature of the program suggests OpenAI will evaluate the fellowship's design and impact before scaling it further.
- OpenAI announced a pilot Safety Fellowship to support independent alignment and safety research.
- The program aims to develop new talent and broaden independent contributions to safety work, per OpenAI’s blog.
- OpenAI describes the fellowship as a pilot, indicating future adjustments based on outcomes.
03 OpenAI move on research
A pilot program to support independent safety and alignment research and develop the next generation of talent
The item stands on its own as a signal from OpenAI Blog.
The main names in the cluster are OpenAI, with the angle leaning on research / safety.
This cluster spans 1 items across 1 sources, so it was treated as a cross-cutting theme rather than a stray item.
- 1 items across 1 sources
- Main names: OpenAI
- Focus: research / safety
What moved around the edges
Hacker News AIで開発ツールの新提案
A Show HN GitHub project (PrismerCloud) demonstrates AI agents that broadcast failure signals to peers so other agents can avoid the same mistakes, with the repository posted as a Show HN item on Hacker News.
Hacker News AIIndian startup Rocket targets consulting reports with an AI platform
TechCrunch reports Rocket launched an AI product that combines strategy, product planning and competitive intelligence to generate consulting‑style reports—positioning the company to offer McKinsey‑like outputs at a lower price than traditional firms.
TechCrunch AIOpenAI lays out a people‑first industrial policy for the AI era
In a blog post titled 'Industrial policy for the Intelligence Age,' OpenAI outlines policy proposals aimed at expanding opportunity, sharing prosperity and building resilient institutions as advanced AI systems scale.
OpenAI BlogWSJ: some workers choose retirement over AI retraining
The Wall Street Journal reports that a portion of workers are opting to retire instead of undertaking AI upskilling or adapting to changing job requirements, a trend highlighted in a Hacker News thread linking to the article.
Hacker News AIHacker News thread requests a browser filter to hide AI posts
An Ask HN post asks for a browser extension that filters out every AI‑related post on Hacker News, reflecting community fatigue with the high volume of AI discussion.
Hacker News AI