llms learn to lie consistently across models
a multi-model study shows that fine-tuned deceptive language models develop early, linearly detectable representations of dishonesty.
author
Baris operates SummarizedData and maintains its sources, publishing rules, and automation. Read the editorial process.
a multi-model study shows that fine-tuned deceptive language models develop early, linearly detectable representations of dishonesty.
nvidia releases cosmos 3, a single open model that generates video, reasons about physics, and predicts actions for robotics and autonomous systems.
this week in ai: developers push back on ai tools, new browsers challenge chrome, and meta plans a wearable pendant.
a look at ai subscription fatigue, browser-based python apps, and the growing pushback against always-on ai.
ai coding agents can amplify distraction, turning quick scripts into abandoned projects and wasting attention, prompting some to consider cutting back on ai use.
box founder aaron levie says ceos are prone to ai psychosis because they are distant from the work needed to create value with ai.
a project runs python asgi web apps entirely in the browser using pyodide and a service worker, removing the need for a backend server except for static files.
chad whitacre steps away from open source and technology, citing ai's disruptive impact and a desire to return to an analog life.
anthropic revenue surges, github copilot billing sparks anger, and meta plans an ai pendant test.
a hands-on look at google's new 24/7 ai assistant shows it handles shopping deals, trip planning, and email summaries well, but lacks key integrations and feels like an unnecessary standalone product.
a look at new browsers using ai, privacy features, and mindfulness tools to challenge chrome and safari.
meta is developing an ai-powered pendant for testing next year, building on its acquisition of ai wearable startup limitless.
github copilot switches from flat subscription to token-based billing, causing cost spikes for some users and sparking debate among developers.
a tool for rendering markdown with live preview and special handling for svg code blocks.
anthropic's annualized revenue projection reaches $47 billion, up from $30 billion in april, as enterprise adoption surges.
ai chip startup groq is reportedly raising $650 million to expand its inference cloud business following a $20 billion licensing and talent deal with nvidia.
a non-developer used google ai studio and gemini to create a quiz about io 2026 announcements.
box founder aaron levie warns of ai psychosis as firms replace workers with agents without understanding their roles.