All activity
Zac Zuoleft a comment
Hi everyone! LightInk is a great example of what happens when a constraint is pushed to the extreme. The project pushes the chip into an ultra-low-power regime: wake stub execution, custom SPI in RTC memory, sub-millisecond display updates, then straight back to deep sleep. That is just a genius engineering move you only find when the constraint is taken seriously enough. And that is what makes...
LightInkE-Ink watch that can run solely on solar power
An open-source, solar-powered E-Ink smartwatch. By re-implementing SPI in the ESP32's RTC wake stub and using a low-quiescent buck-boost converter, it reduces active boot time to under 1ms, delivering up to 10 months of battery life on a 100mAh charge.
LightInkE-Ink watch that can run solely on solar power
Zac Zuoleft a comment
Hi everyone! Gemini App keeps choosing to go broad. This update brings the output much closer to everyday workflows: Docs, Sheets, Slides, PDFs, Word, Excel, CSV, Markdown, LaTeX, and more can now be generated directly inside the chat. No more copying a good answer, pasting it elsewhere, and fixing the formatting by hand. That says a lot about Gemini app’s direction. It is trying to be a wider...
File Generation in GeminiGenerate production-ready files directly in your chat
Zac Zuoleft a comment
Hi everyone! After Samsung moved away from S Pen support on the Galaxy Z Fold 7, and its new tri-fold also launched without it, Motorola’s Razr Fold lands in a pretty interesting spot. A large book-style foldable that still leans into pen-enabled productivity. The Moto Pen Ultra is sold separately, but the positioning is obvious — a bigger canvas for notes, annotations, screenshots, and more...
Motorola Razr FoldA foldable phone built for pen-first productivity
Motorola Razr Fold is Motorola’s first book-style foldable, with an 8.1-inch inner display, Snapdragon 8 Gen 5, 6000mAh battery, 80W charging, a DXOMARK Gold-rated foldable camera system, and Moto Pen Ultra support for notes, annotations, and big-screen productivity.
Motorola Razr FoldA foldable phone built for pen-first productivity
Gemini can now generate downloadable files directly in chat, including Google Docs, Sheets, Slides, PDFs, Word, Excel, CSV, LaTeX, Markdown, TXT, and RTF. Go from prompt to ready-to-share file without copying, pasting, or reformatting.
File Generation in GeminiGenerate production-ready files directly in your chat
Gro’s new Content Search + Social Signal Monitoring helps you find what your buyers are already talking about, identify the authors worth engaging, and then automatically monitor for new high-intent posts so you can respond the moment intent shows up (you see the post).
When a matching post appears, Gro can trigger your playbook instantly: alert you, comment, send a connection request, and draft/send email/message — all from one workflow. From content → to leads → to monitoring → to action.
Gro v2Spot signals, trigger outreach - turn posts into pipeline
Zac Zuoleft a comment
Hi everyone! Earlier this April, a mysterious model called HappyHorse suddenly showed up on Arena and jumped straight to the top of the video leaderboards. Now the reveal is here: it comes from Alibaba ATH, and apparently HappyHorse is not just a codename — it is the actual product name. Honestly, pretty fun. The model already looks strong across the core video tasks: text-to-video,...
Happy HorseTop-tier AI video generation and editing from Alibaba ATH
HappyHorse is an AI video creation platform powered by Alibaba ATH’s top-ranking video model. It lets creators generate and edit videos across text-to-video, image-to-video, and video editing workflows in one simple product.
Happy HorseTop-tier AI video generation and editing from Alibaba ATH
Zac Zuoleft a comment
Hi everyone! I like @Odyssey’s framing of this as pretrained physical intelligence. The model learns how scenes evolve, how objects move, how interactions play out, and how the world holds together over time. Odyssey-2 Max is still early, but this direction makes the robotics, gaming, simulation, and interactive systems angle much easier to understand.
Odyssey-2 MaxPhysical accuracy takes a leap in world models
Odyssey-2 Max is Odyssey’s largest general-purpose world model yet, built for real-time interactive simulation. It uses autoregressive next-state prediction to improve physical accuracy, long-horizon stability, and open-ended interaction across worlds that evolve with user actions.
Odyssey-2 MaxPhysical accuracy takes a leap in world models
Zac Zuoleft a comment
Hi everyone! The long-awaited DeepSeek V4 is finally here, and the message is simple: 1M context is becoming normal. V4-Pro is the flagship model, with stronger agentic coding, world knowledge, and reasoning. V4-Flash is the fast, efficient version for more economical use. Both models support 1M context and are available through API today, with open weights already released. DeepSeek’s real...
DeepSeek-V4The open-source era of 1M context intelligence
Zac Zuoleft a comment
Hi everyone! This is a milestone release from Tencent. After rebranding their Hunyuan foundation models to the "Hy" series, pronounced like "hi", this is the first model shipped since they rebuilt their pre-training and RL infrastructure in February 2026. The model itself is a 295B MoE with 21B active parameters, 256K context, and three reasoning modes for different latency / depth tradeoffs....
Hy3 previewThe first step in rebuilding Tencent Hunyuan model
DeepSeek-V4 Preview is a new series of highly efficient MoE language models, featuring V4-Pro (1.6T params) and V4-Flash (284B params). Both models support a 1 million token context window by default, utilizing a novel hybrid attention architecture to drastically reduce compute and memory costs.
DeepSeek-V4The open-source era of 1M context intelligence
Hy3 Preview is a 295B-parameter MoE model (21B active) from Tencent. Built on a completely rebuilt RL infrastructure, it delivers best-in-class cost efficiency and significant improvements in complex reasoning, instruction following, and agentic coding tasks.
Hy3 previewThe first step in rebuilding Tencent Hunyuan model
Zac Zuoleft a comment
Hi everyone! This is a dense model, not an MoE, and that matters. Dense models often show unusually strong intelligence density for their size, and Qwen3.6-27B is a very good example of that. At 27B, it already pushes past Qwen3.6-35B-A3B on a number of key coding and reasoning tasks, and more importantly, it beats the previous open-source flagship Qwen3.5-397B-A17B across all major coding...
Qwen3.6-27BThe sweet-spot open dense model for coding agents
Zac Zuoleft a comment
Hi everyone! Xiaomi MiMo ships new models at a pace that’s honestly astonishing. MiMo-V2.5-Pro is their strongest yet — big jumps in agentic coding, long-horizon tasks, and real-world reliability. The lighter V2.5 brings native multimodal (image/audio/video) on top of that while staying very token-efficient. MiMo is clearly accelerating to become one of the fastest-moving players in the...
MiMo-V2.5 & ProFrontier agent capability with better token efficiency
Qwen3.6-27B is a fully open-source dense model that punches way above its weight. Surpassing the previous 397B MoE flagship in agentic coding, it supports multimodal reasoning and thinking modes while remaining perfectly sized for local self-hosting.
Qwen3.6-27BThe sweet-spot open dense model for coding agents