Trending or useful GitHub repos / working software
27 pieces
5 recent pieces
On 2 October 2026, Supabase’s developer day walked Compute, schema-in-git, native local, app MCP, ops agents, OrioleDB, Multigres, and Turso under three words. Capital news from that morning sits on Frontier’s earlier Dispatch.
Read the piece22 stories
Code retrieval jumps 9.92 points. Google's own audio chart plots a sub-1B rival higher, and that rival bars commercial use.
The hosted API answers noul, choice and score questions against text, JSON or images. Docs list a 262,144 token input limit and 10 requests per second per organization.
The Linux SDK gives Muse system.run, file.read and file.write as the install account. Tokens are personal and non-commercial, and Meta says the kit is not a supported product.
The 27B and 9B models return probabilities for typed questions in one pass. Cloudflare prices Clef at $0.24 per million input tokens and says median latency is 209 ms.
The 3 October release needs one H200 or two A100 80 GB cards plus a vLLM plugin. Aleph Alpha's blog and model card give different German shares of the training mix.
The 1.9B model answers yes or no, choice and score questions in one pass, with a median 115 ms on an RTX 3090. Its maintainers rate it an experiment.
Black Forest Labs released a fine-tunable world action model that maps camera frames and text to action chunks. It reports 42.92% on RoboLab-120. The weights use the FLUX Kommunity License.
Local k3s runtime, ICON real-time control, FoundationPose perception, and OMTS machine-tending reference under ROS 2 Lyrical Luth.
One TypeSafe request picks operation and target. A small LLM types only on TYPE_TEXT. MIT license. Repo docs report a 7.1s Google Flights demo.
Declare Task, Workspace, Gateway, and Model as ax.io/v1alpha1 manifests. AX sandboxes the run, fences egress, and can suspend or SSH a live agent.
OpenAI's first-results post (25 August 2026) and Broadcom unveil put measured first-party silicon on the record. This Stack piece inventories the architecture claims, maturity, and pick-vs against merchant GPUs.
NVIDIA's 16 September 2026 preview submission puts Vera Rubin NVL72 on DeepSeek-R1 and Qwen3-VL. This article covers the benchmark results and software stack, separate from earlier Vera Rubin production news.
NVIDIA's 9 September 2026 toolkit release extends CUDA to Windows on Arm, previews Rubin (sm_107), modernizes shared-GPU control, and changes memory-management defaults on coherent platforms.
When to rely on native AGENTS.md load, when to keep @AGENTS.md in CLAUDE.md, and why Bedrock, Vertex and Foundry sessions need 2.1.281 for direct load.
NVIDIA opened native GPU kernels in Rust on 8 September 2026. cuda-oxide is SIMT on pinned nightly. cutile-rs is Tile on stable Rust. Pick by control need, not by logo.
NVIDIA announced a $12,930,300,000 acquisition of Hugging Face. The 8-K adds about $11.9 billion payable to stockholders, up to $1 billion in retention equity, close expected in the first half of 2027, and required regulatory approvals. Hub neutrality is NVIDIA's public promise, not a close condition printed in the filing.
OpenAI, Google, and Anthropic all ship frontier cyber capability behind trusted-access walls. The public model is trained to refuse the interesting work. The gated program is where that work is supposed to live.
OpenAI says Astra meets Critical cybersecurity and is harder to read than Sol. Pachocki says no lab has solved monitoring enough to keep scaling at maximum speed. Microsoft's draft Code of Conduct bans concealed reasoning for MAI models only.
Apache-2.0 serving framework. v0.5.14 landed 26 June 2026 with DeepSeek-V4 and Waterfill/LPLB load balancing.
27 March 2026 checkpoints on Hugging Face `facebook/sam3.1`. Meta reports ~7x at 128 objects on one H100 versus SAM 3. Weights gated. GitHub stars not published.
Google Research, 13 January 2026. Weights are not Apache-2.0. Clinical diagnosis is out of intended use. Internal accuracy deltas are Google's.
New releases, then the constraint.
