#video #lecture by Peter Alvaro of UCSC at Strange Loop about disorderly systems and “I see what you mean: what a tiny language can teach us about gigantic systems” #toread
Rustyaxe’s code, including #ham #radio #communications software
Cognitect got acqui-hired by Nubank in 02020, but continue to maintain #Clojure and (their proprietary database) Datomic. #Lisp #databases
#Spymarks are watermarks used to spy on you
William MacAskill explains in 02014 where #EA came from as a term: in October 02011 he asked the 80,000 Hours team how to rebrand (rather than “do-gooder”) and in December they voted for “Center for Effective Altruism”. The voters were, as far as he could remember three years later, himself (Will MacAskill (then ‘Crouch’)), Toby Ord, Nick Beckstead, Michelle Hutchinson, Holly Morgan, Mark Lee, Tom Ash, Matt Wage, Ben Todd, Tom Rowlands, Niel Bowerman, Robbie Shade, Matt Gibb, Richard Batty, Sally Murray, Rob Gledhill, and Andreas Mogensen. #history
#Video of Andrew Kelley's #lecture at #SXSW this year saying that (specifically for #AI) “if you don’t control the source code, you are not in control,” claiming AI agents are under the control of AI companies, which is of course true. There’s a section about disliking AI but using it anyway, and a section explaining how #Anubis is killing the open Web.
Very clear statement about the value of #free-software: “It primarily matters whether high-quality free software is available to users, providing competition that limits the degree to which proprietary software can get away with being coercive.” #toread
#video #toread #formina of a luxury train trensolidario.org that goes 1200km from Haedo to Senillosa, Neuquén, in 50 hours, with a misleading title that says it goes to Patagonia instead. The fare is 20kg of non-perishable food. The car is a wood-paneled “Pullman” car from the mid-01950s that has recently been restored.
#video #toread about the aestheticization of #politics in modern #fascism
#video #toread about #Freenet and services without servers. #decentralization
#humor #video about how the #EU impedes technical innovation and how it’s hard to start startups there
#video #lecture by John #Ousterhout about how his new RPC protocol #Homa improves tail #latency compared to TCP/IP and RDMA (RoCE)
#video by #Lukes-Dev-Lab testing a new #Qwen finetune, a DavidAU’s 4-bit quantization of Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NEO-CODER-MAX-MTP-GGUF, which performs amazingly at some software development tasks, especially front-end web dev, although it fails at #Godot development. #LLM #AI #neural-networks
#news that #NASA has canceled its #Mars sample return mission
#video review of #Anthropic’s #Claude Opus 5.5, which came out today. He had it write some 3-D games from a single prompt, using threejs, #Godot, and just plain C++ with no dependencies. He was not able to get it to drive a simple robot arm to push a toy car across a mat, though. “This model slaps.” #AI #LLMs #neural-networks
[Submitted on 8 Jul 2026 (v1), last revised 6 Sep 2026 (this version, v2)] “Recursive Self-Improvement in #AI: From Bounded Self-Refinement to Autonomous Research Loops”, by Mingguang Chen, Licheng Wang, Bo Qu. #paper. “We survey 1,250 arXiv papers (2024-2026) along two axes: what the system improves -- its behavior in deployment, its policy through training, its evaluator, or the research process itself -- and the degree of loop closure (human-in-the-loop to fully closed). The taxonomy separates bounded self-refinement -- convergent, evaluable, and industrial practice -- from open-ended recursive self-improvement (RSI), which remains bounded by grounding requirements, collapse dynamics, and compute constraints on every measured axis. Its distinctive feature is a dedicated category for self-evaluation: every improvement loop is a claim that some signal can substitute for human judgment.” 44 pages. #toread
#Breaking-Taps #video about how he got silicon #electronics fabricated (through #WaferSpace #MPW, not a multi-project chip). He settled on a MOV machine (transport-triggered architecture) called BTTAP, designed it in the Spade #HDL, synthesized it with Yosys and ABC, and simulated it for SKY130. Includes animations of some analog simulations. He tested his RTL (?) in cocotb instead of Verilator and viewed his waveforms with Surfer instead of gtkwave. Then he used LibreLane, which wafer.space provides as a #Nix Flake, to get from a netlist to GDSII. Finally wafer.space got it fabbed through GlobalFoundries’ GF180 process, which I didn’t know had an open-source PDK, in only 3-4 months. Wafer.space got some of the chips packaged as COB with transparent epoxy. He got about 1000 chips. From looking at MPW pricing in the past, I’d estimated that this cost him about US$2500, but I think it was actually US$5000. wafer.space GF180MCU run 3 is in December, early-bird price for a quarter-size chip (1.94 × 2.53 mm) of only US$2000.