Today we are shipping Pi 1.0, a hardened, minimal, extensible agent harness, alongside Pi Durable, a new experimental substrate for long-running agentic applications.

We are introducing Clef and Clef-flash, open-source decision models hosted on Workers AI for high-speed classification and agentic workflows. Also launching: a new reinforcement learning platform that allows developers to fine-tune decision models using their own data.

This ticket is something like the master ticket to coordinate development on an iOS port of StreetComplete. It replaces #1892 which also included a lot of discussion and research / observation work...

Git 3.0 will make SHA-256 the new default content hashing algorithm and it will be an incomprehensibly expensive and ultimately valueless and avoidable global nightmare.

Pi Durable is a new experimental package for long-running, durable, and malleable agents that can run anywhere. A tour of what we built and why.

Reminder: The "border exemption" to your rights is real.

Get it while it’s hot

Micron CEO Sanjay Mehrotra says memory and storage supply will likely be much tighter in 2027 and 2028 than it was in 2026. He made the remarks during Micron's fiscal Q4 2026 earnings call on September 30, where the company reported record quarterly revenue of $54.2 billion. Here's what he...

turbopuffer is pushing the frontier of search. To do that, we have to fundamentally redesign our storage architecture so the vector index is no longer primary.

Previously, I started a new Play Store account because I wanted to make video games for Android. But I never got time to finish my game. As a result, my personal play store account was closed by @Google because of inactivity.

Now, I developed a new app, let's publish it on play store. But I canno…

Back in 2025, we posted about ESPARGOS, a phased array of many patch antennas, each connected to an ESP32 WiFi microcontroller. ESPARGOS could be used to determine the direction of arrival for WiFi signals and create a live augmented reality heatmap. Recently, the ESPARGOS team has made an exciting discovery. They found that several ESP32 chips have an undocumented feature that lets the firmware bypass the fixed WiFi and Bluetooth functionality and instead capture raw IQ baseband samples. As a result, several ESP32 models can now be used as an internal SDR covering 2.2–2.7 GHz, plus 4.8–6.0 GHz on the

Magent Forensics, the owner of the GrayKey phone unlocking tool, says it can bypass an iPhone rebooting feature that was locking cops out.

Cloudflare K2 is a serverless event streaming service built directly on top of R2 object storage for high-scale data movement and long-term retention. By decoupling producers and consumers at the edge, K2 enables durable, ordered log streams without the operational overhead of traditional broker clusters.

My last post on the Rust compiler’s performance was two months ago and a lot has happened since then.

The full 1996 dial-up experience: the handshake, the wait, someone picking up the phone. Sound on.

A Vulkan reimplementation of NVIDIA's DLSS 5 Neural Rendering network, bit-exact against the original. - maanHimself/OpenDLSS-NR

A curated gallery of generative patterns. Discover, customize, and download unique algorithmic art.

Rescuing a field from disappearing

The probe adds to the mounting scrutiny that OpenAI and Anthropic have been facing over their safety practices following the Hugging Face hack.

Companies form strategic partnership to dramatically advance and accelerate semiconductor innovation News Highlights: Companies sign multi-year agreement to collaborate as preferred partners to...

We introduce Context Language Models (CLMs), language models that natively manage their own context. We implement this by treating the context as a file and allowing the model to make unrestricted updates to this file. This allows the model to learn what is most important to maintain in context, and naturally extends to multi-agent systems where multiple agent contexts coexist as files. Building CLMs zero-shot with existing models outperforms SOTA context management strategies across a variety of tasks: 11.4% higher accuracy with 21.5% fewer FLOPs on BrowseComp-Plus, 5% higher scores with 59% fewer FLOPs on 12-hour EdgeBench, and 65% greater improvement with the same compute on a 24-hour multi-repository agent-swarm task. Moreover, by shifting context management from external harness control to intrinsic model behavior, CLMs naturally enable both in-context and parametric learning of context-management strategies. We show that CLMs can be steered with natural-language instructions evolved through a standard skill-optimization loop, improving held-out accuracy by up to 35.9 points on a context-management task while reducing compute. We also introduce an online reinforcement learning method for CLMs, improving Qwen3.5-9B performance on BrowseComp-Plus by 47.6% while using 12% fewer FLOPs. Finally, we co-design Suffix Cache Reuse for CLM serving, further reducing server-side compute by 35% relative to standard SGLang at matched performance.

Hey David, Gayani here from Figma. You're right that our remote MCP server only accepts clients on our supported list, and Pi isn't on it yet. You can see the current list in our MCP catalog at https://t.co/T0hj2nTIWa. If you'd like Pi considered for a future addition, please fill out this form: ht…