Tracked 227 · Curated 15
Artist SHL0MS posted a real Claude Monet painting, claiming it was AI-generated, which prompted widespread criticism calling the 'AI image' emotionless and lacking depth. The incident revealed a strong negative bias against AI art among the public, despite research indicating people actually prefer AI-generated works.
In a special episode, Yaroslav Azhnyuk, founder of The Fourth Law, discusses the escalating AI-driven drone warfare, contrasting Ukraine's 4 million FPV drones last year with China's potential 4 billion. He argues the West is planning for the last war, lagging in drone technology, tactics, and geopolitics, with the gap widening due to China's manufacturing advantage and Western defense readiness issues.
The new open-source plugin, claude-smart, aims to enhance Claude Code by introducing 'smart memory'. Instead of just storing and retrieving information, it learns from past experiences to 'self-improve', making Claude Code more effective in future interactions.
Rubric-based On-policy Distillation (ROPD) is a new framework for distilling Large Language Models (LLMs) that replaces teacher logits with semantic rubrics, enabling black-box distillation. ROPD achieves 10x sample efficiency gains over logit-based methods and works with proprietary API-only models. It induces prompt-specific rubrics from teacher-student contrasts and uses them to score rollouts for on-policy optimization, making it compatible with any black-box teacher.
BuzzFeed CEO Jonah Peretti explained the sale of 52% of the company for $120 million to Byron Allen, describing it as a "new lease on life" that helps the company overcome cash shortage risks. Peretti is stepping down as CEO to become president of BuzzFeed AI, with Byron Allen taking over as CEO.
The article highlights a shift in AI programming paradigms. Previously, developers wrote code to call Large Language Models (LLMs). Now, the focus is on writing prompts and skill files to instruct LLMs to execute code. The author suggests the future of AI programming remains unwritten.
Planned research by the University of Washington to equip preschool teachers with cameras for AI training has been shelved due to parental backlash. The 'opt-out' nature of the program raised concerns over informed consent and data privacy. While researchers aimed to assess classroom interaction quality, parents and experts questioned data sharing and long-term usage.
A predicted 2026 trend involves friends casually showing each other their "vibe-coded" life/work dashboards and AI setups, likened to teenagers showing off their Magic: The Gathering binders.
Former Google CEO Eric Schmidt was met with boos from students at the University of Arizona commencement when his speech turned to AI. Schmidt acknowledged the anxieties about job loss, environmental issues, and inheriting a 'mess' as 'rational,' but the lecture was largely drowned out by the negative reaction.
A user expressed that for writing tasks, Opus 4.6 is the best model, while for translation, Gemini 3.1 Pro is the best. The user is soliciting feedback on when others choose models other than Codex and what frustrations they have, to aid in improving the next generation model.
AI Engineer (AIE) is set to launch in India soon. This move is supported by swyx and is expected to happen next year with assistance from Sanjeed_i and Udayan_w.
The article argues that the situation will be dire if a strong American open-source champion doesn't emerge soon, especially if major labs succeed in banning open models on national security grounds. Project Tapestry is presented as a potential salvation for the open-source movement.
The FBI is seeking to purchase nationwide access to automated license plate reader (ALPR) systems, which could allow the agency to track vehicle and people's movements across the country without a warrant, according to procurement records. This move comes amid increasing pushback against ALPR technology. The FBI plans to spend up to $36 million for access across major US regions and territories, aiming to acquire comprehensive data.
AIE criticizes the unauthorized and uncredited reposting of its video content by 'influence operators,' attributing it to excessive 'hype lingo.' For example, a two-hour workshop on Anthropic's Claude Agent SDK was reposted without crediting the original speakers. AIE urges users to report such instances to the original account @Jouhatsu_ai.