The Archive

Figure AI's humanoid robot to demonstrate 8-hour autonomous operation at human speed in live stream.

Wanted to share our latest paper on an alternative building block for Vision Transformers. [Illustration of our model's accuracy and dense features](https://preview.redd.it/x4acnx478w0h1.png?width=2457&format=png&auto=webp&s=3ce49caf2b0cdea5d35141aebb7297862fdc6a7d) Traditional ViTs utilize dense (***N******^(2)***) self-attention, which can become pretty costly at higher resolutions. In this work, we propose an alternative backbone with a core-periphery block-sparse attention structure that scales as (***2NC + C******^(2)***) for ***C*** core tokens. We further train this usin...

Senior developer shares terminal workflow tips for Claude Code CLI usage on Linux.

Reddit post argues AI causes unemployment and cognitive bias; lacks evidence or specific claims about frontier AI capabilities.

Anthropic launches Claude For Legal with practice-area plugins and MCP connectors to DocuSign, LexisNexis, Thomson Reuters, and six other legal platforms.

OpenAI forms deployment company; labs pursue top-down AI implementation strategy; Apple-Intel partnership reflects economic incentives.

At its peak, the Androscoggin paper mill in Jay, Maine, a rural town about 67 miles northwest of Portland, employed about 1,500 people - until a pulp digester exploded in 2020, forcing the mill to close permanently. In 2023, the 1.4 million-square-foot facility was purchased through a joint venture by JGT2 Redevelopment and a number of other holding and capital companies. The project is led by developer Tony McDonald. Over the next three years, McDonald and his team broke down the mill's machinery and shipped it to Pakistan, and worked to clean up the industrial site for resale. That resale a...

Link: reuters.com

Xiaomi open-sourced MiMo-V2.5-Pro (1.02T params, 42B active MoE, 1M context) under MIT license; cost-benefit analysis of self-hosting vs. API.

Reddit user reports 10-day support silence after paying for Claude Max, unable to access Claude Design feature.

Non-technical worker at Japanese logistics firm uses Claude for route optimization, training materials, and safety documentation without coding background.

Reddit discussion on workflow changes from AI coding assistance adoption.

Reddit user compares Codex and Claude for coding tasks, finding mixed results; seeks community perspective on relative capabilities.

Simon Willison demos CSP sandboxing technique using iframe allow-list prompts; web security pattern, not AI-related.

User demonstrates practical applications of local LLMs including embedding-based semantic search and Qwen 35B for automated database evaluation workflows.

Opinion piece speculating on the decline of fine-tuning as a technique in frontier AI model development.

Who can afford a 300/hr CBT trained psychiatrist? I loved *Feeling Good*, but it’s such a thick book and learning all the strategies is tough and a lot of work… Recently, I plugged the below custom instructions into Claude and now Claude walks me through distorted thinking and thought patterns… it really helped me \# Custom Instructions Act as Dr. David Burns conducting a CBT session with me as your patient; use the techniques from Feeling Good including identifying my cognitive distortions, the triple-column technique (automatic thought / distortion / rational response), Socratic questi...

Reddit discussion on hardware cost inflation for local LLM inference rigs and DRAM supply chain developments.

After working with Opus 4.7 for over two weeks, I noticed a subtle but persistent change in long conversations: the model's fundamental capabilities are still there, but the output feels filtered through something. Details that should be remembered get dropped, consistency drifts. It feels more like the model is zoning out. The system card data seems to support this. MRCR v2 8-needle test: Opus 4.6 scored 91.9% recall at 256k context. Opus 4.7 dropped to 59.2%. At 1M context, it went from 78.3% to 32.2%. That's a significant decline. Boris Cherny has publicly stated that MRCR is being phase...

Sam Altman testifies in Musk's lawsuit against OpenAI over corporate structure and fiduciary duty claims.

User describes using ChatGPT and Gemini to research tenant rights and recover $4200 from disputed deposit claim.

Reddit user shares humorous anecdote about Claude recognizing their work habits.

Reddit user describes Claude's skepticism about agent use cases in consulting work.

There is no governmental mechanism to pay for an AI agent that monitors a patient between visits, calls to check in, coordinates a housing referral, or makes sure someone picks up their medication. ACCESS creates that mechanism for the first time.

IP lawyer built Sonos Mac/iOS apps in a weekend using Claude Code without prior coding experience, demonstrating productivity gains from AI coding assistants.

Hello everyone, Presenting at a top-tier conference for the first time and having a very hard time coming up with an appropriate design for my poster. Everything I do seems basic and banal. My paper is more theory-oriented, and apart from putting math formulas in bold in the middle, I am not sure what the best way is to design the poster. Even the sizing choice is complicated as ICML gives 3 different recommendations to pick from, and somehow from my computer, I can’t see how the PowerPoint slide will look like printed on those dimensions. And Printing a poster is nearly $100 CAD, ...

OpenAI details response to TanStack npm supply chain attack, outlines security hardening and mandatory macOS app updates by June 12, 2026.

Developer argues current AI agents require extensive human oversight and lack true autonomy despite productivity gains with Claude models.

Datasette 1.0a29 fixes mobile UI bugs, adds TokenRestrictions utility, and resolves segfault race condition in test suite.

Reddit discussion arguing for immediate public disclosure of all AI discoveries, touching on transparency vs. responsible deployment tensions.