The Archive

PhD student reports 4% accuracy gap when reproducing computer vision paper baseline; raises reproducibility concerns common in published ML research.

Reddit post with no content; insufficient information to assess.

Boris Cherny, creator of Claude Code, discusses loops as a future direction in a podcast appearance.

Unexpected email to wake up to but I am here for it! Model agnostic tools are the way! This is huge!

Community member merges Qwen3.6 chat template fixes from froggeric and allanchan339 using Claude Opus.

Staffers at Google DeepMind's headquarters have voted to unionize in an effort to prevent the AI firm's technology from being used by Israel and the US military. In a letter to Google management on Tuesday, employees requested that the Communication Workers Union (CWU) and Unite the Union be recognized as joint representatives, with 98 percent of CWU members at DeepMind voting in support of the move. "We don't want our AI models complicit in violations of international law, but they already are aiding Israel's genocide of Palestinians," an unnamed DeepMind employee said in a statement shared ...

Dual RTX 3090 setup draws ~760W under LLM inference, 90W idle; practical hardware benchmark for on-premises deployment.

Stratechery analysis: Amazon lagged in AI training but positioned for inference dominance through sustained infrastructure investment.

OpenAI releases GPT-5.5 Instant system card detailing model capabilities, limitations, and safety properties.

OpenAI releases MRC (Multipath Reliable Connection), an OCP networking protocol for resilience and performance in large-scale AI training clusters.

OpenAI releases GPT-5.5 Instant as ChatGPT's default model with improved accuracy, reduced hallucinations, and personalization controls.

Backend developer reports organizational burnout after Claude adoption enabled management to reduce headcount without reducing workload.

Developer built hand-tracking web app with Claude assistance; dual-hand detection with visual effects.

Every few centuries, changes in how information moves reshape how societies govern themselves. The printing press spread vernacular literacy, helping give rise to the Reformation and, eventually, representative government. The telegraph made it possible to administer vast nations like the US, accelerating the growth of the modern bureaucratic state. Broadcast media created shared national audiences,…

Link:

Humorous video compilation of ChatGPT generating incorrect responses.

Anthropic released Claude for Creative Work with nine MCP-native connectors including Blender, enabling persistent in-app context and direct action execution.

User reports hitting Claude Pro daily token limits faster than usual; anecdotal observation of rate-limit behavior change.

cred: fabianstelzer

vibevoice.cpp: C++ ggml port of Microsoft VibeVoice enables TTS and long-form ASR with diarization on CPU/CUDA/Metal/Vulkan without Python.

User reports account suspension from Claude after linking Spotify integration; anecdotal complaint without confirmation of cause.

I built Claude version of pets. Here is the repo if you want to try: [https://github.com/alvinunreal/claude-pets](https://github.com/alvinunreal/claude-pets) You can find more pets over here: [https://openpets.dev](https://openpets.dev)

DeepSeek V4 Pro matches GPT-5.2 on FoodTruck Bench agentic task; first Chinese frontier model, 17× cheaper, within 10 weeks of GPT-5.2 release.

User automated a 5-step lead enrichment pipeline using Claude with MCPs, replacing Apollo, PDL, and manual HubSpot loading in a single workflow.

MTP format support coming to llama.cpp; DeepSeekv3, Qwen3.5, GLM4.5, and other models compatible pending native weights.

Qwen 27B FP8 achieves 80 TPS with 200k token BF16 KV cache on RTX 5000 PRO 48GB, reducing quantization artifacts vs. 24GB quantized baselines.

US GUARD Act advances in Senate with mandatory age verification for AI chatbots, framed as child safety but viewed as broader regulatory friction.

Anonymous text-to-image model 'Peanut' ranks #8 on Artificial Analysis arena; open-weights release promised to compete with FLUX and Qwen models.

User reports Claude Design service outage with data loss; service restored after ~1 hour.

The Nvidia CEO seems to feel that claims of AI's job-killing potential have been greatly exaggerated.