Daily Tech News - 2026-09-26

Today’s tech pulse runs from silicon’s smallest details to AI’s expanding role in how software is built, tested, and understood.

At the systems level, developers are confronting both old mysteries and new hardware. A lost atomic update on LoongArch’s LA664 turned a routine Debian packaging job into an infinite-loop investigation, while a Linux AF_ALG vulnerability retrospective explains how a local privilege-escalation flaw earned a $113,337 bounty. Intel’s Panther Lake teardown highlights gate-all-around transistors, backside power delivery, and advanced packaging; meanwhile, reverse-engineering the vintage 8087 reveals that its tangent algorithm was more sophisticated than a straightforward CORDIC implementation. Rust’s SIMD ecosystem is also maturing, and a proposed Linux 7.4 memory-management change is producing eye-catching—if highly benchmark-specific—results.

AI agents increasingly form their own software layer. Emerging standards aim to connect models with tools, remote agents, organizational knowledge, users, and IDEs. Floci brings AWS, Azure, GCP, and OCI emulation into a fast local workflow, while Drawgent places Claude Code, Codex, or OpenCode on a live Excalidraw canvas. Reladraw offers another take on programmable diagrams, giving authors explicit control over placement. Experiments with code-review graphs, Jev-style interfaces, locally hosted decision models, and LLM watermarking show that agent infrastructure is advancing faster than consensus about its reliability. The newly launched The Code Is Lying captures the central tension: as generating code becomes cheaper, verifying and understanding it becomes more valuable.

Elsewhere, engineers are returning to fundamentals. Guidance on repeatable server benchmarks, layered timeout defaults, HTTP 413 errors, “parse, don’t validate,” and the tradeoffs created by common system-design fixes all reinforce the same lesson: abstractions help, but production failures still emerge at their boundaries. Even an old Steam Link reborn with NixOS fits today’s theme—curiosity remains one of computing’s most productive tools.

Featured Articles

The lost atomic update on LoongArch LA664

One CPU Atomic Instruction, One Packaging Infinite Loop: The Story of the Lost Update on LA664¶ TL;DR¶ In February 2026, Wang Miao ran into something strange while packaging normaliz for Debian on a L...

Keywords: cpu atomic, instruction fault, loop investigated, problematic atomic, instruction atomic, atomic instruction, fails atomic, heap corruption, openmp process, atomically incremented
Source: jia.je

Floci: Locally emulating any cloud service

Any cloud. Locally. Run AWS, Azure, GCP, and OCI locally in milliseconds: the fast, credential-free loop your team and its AI agents need to ship faster. Run AWS, Azure, GCP, and OCI locally in millis...

Keywords: cloud code, azure gcp, aws azure, run cloud, cloud run, write cloud, local cloud, cloud locally, cloud services, cloud emulators
Source: floci.io

Infecting the Steam Link with NixOS

While rummaging through my closet the other day I discovered a Steam Link I had bought on flash sale way back in 2018 still dutifully humming away all these years later. It occured to me that having a...

Keywords: nixosconfigurations steamlink, kexec steamlink, kernel nixos, bootable nixos, linux steamlink, kexec nixos, kexec boot, self nixosconfigurations, boot kernels, kernel kexec
Source: feyor.sh

A single function Jev-like wrapper for LLMs, including vision models

I was intrigued by Jev and the self-hostable projects appearing around it, such as OpenJev and SemIf. Reading about them introduced me to a neat trick: reading an LLM's token probabilities. Apparently...

Keywords: chat completions, max_completion_tokens logprobs, token alternatives, alternative tokens, request max_completion_tokens, tokens, max_completion_tokens, jev request, chat logprobs, llm token
Source: allanrbo.blogspot.com

IBM Bob with code-review-graph: A Benchmark on Open Liberty

Code graphs sound like a very good match for coding agents. And there is rarly a day where not another one of these graph frameworks passes my social streams. The idea is simple. Parse the repository...

Keywords: graph agent, code graphs, code graph, graph frameworks, repository graph, coding agents, graph stores, graph assisted, graph navigation, graph source
Source: the-main-thread.com

Joy & Curiosity #101

Last week I was on Matt Swanson’s podcast and we ended up sharing thoughts and vague predictions about programming languages and frameworks. Matt said that he’d be going to Rails World the week after...

Keywords: shared abstractions, abstractions worth, say rails, frameworks longer, languages frameworks, said rails, frameworks past, rails instead, talk rails, worth framework
Source: registerspill.thorstenball.com

The Emerging Standards Behind AI Agents

To do useful work, an agent must connect to a model, tools, remote agents, users, organization-specific instructions and knowledge, operational systems, and, for coding work, an IDE. Agents are becomi...

Keywords: agents tools, agents workflows, agentic workflows, agents connect, connects agent, agent plugins, agent tool, agent connect, ide agents, agent schemas
Source: generativeprogrammer.com

The state of SIMD in Rust in 2026

A lot of progress was made since last year, and I made some of it! After last year's survey I started contributing to the SIMD library that seemed the most promising. One thing led to another, and now...

Keywords: implementations simd, useful simd, simd library, using simd, simd operation, simd hardware, designed simd, simd abstraction, simd directly, usingstd simd
Source: shnatsel.github.io

Google researcher explains why Windows NT “puts Linux to shame,” and imagines an alternate history where it won

A Google researcher who spent four years as a reverse engineer at Microsoft has reopened one of computing’s oldest arguments, and she thinks Linux is on the losing side. Laurie Kirk, who joined Google...

Keywords: kernel windows, nt kernel, nt os, argues windows, os nt, isn windows, government windows, windows nt, os windows, os define
Source: windowslatest.com

Rusty thoughts on "Parse, don't validate"

Like many programmers, I find Alexis King's Parse, don't validate article fascinating, because it gives a name to an idiom that seems familiar and important - one I've observed and used in the past wi...

Keywords: rust defines, rust programming, code rust, idiomatic rust, example rust, rust rewrite, rust universally, exist rust, integers rust, rust generic
Source: eli.thegreenplace.net

Tuning a Server for Benchmarking

Tuning a Server for Benchmarking Taming OS and hardware noise for repeatable benchmarks.Optimizing code starts with measuring it, and a measurement is only useful if it is repeatable: a 2% improvement...

Keywords: tuning benchmarking, benchmarking tuning, repeatable benchmarks, server benchmarking, benchmarks optimizing, performance benchmark, benchmark tune, performance clocks, benchmarking taming, core benchmark
Source: david.alvarezrosa.com

Make Claude your assistant in excalidraw

drawgent: your coding agent on a live Excalidraw canvas# drawgent connects your own Claude Code, Codex or opencode (your install, login, config and repo) to an Excalidraw whiteboard. Ask for a diagram...

Keywords: drawgent setup, drawgent config, automatically drawgent, commands drawgent, config drawgent, canvas drawgent, add drawgent, browser drawgent, npm drawgent, sudo drawgent
Source: tangled.org

Intel Panther Lake Teardown

Panther Lake debuts the first commercial implementation of backside power delivery (BSPDN), introduces Intel’s first iteration of gate-all-around (GAA) transistors, and showcases their advanced packag...

Keywords: intel gate, fabrication gate, gaafets gate, gate technology, intel foundry, foundry intel, gaafet designs, intel manufacturing, intel tech, lake intel
Source: newsletter.semianalysis.com

How I Found a $113,337 AF_ALG Linux Local Privilege Escalation Before Copy Fail

In 2025, I found an AF_ALG vulnerability in the Linux kernel that allowed an ordinary user to escalate privileges to root. This is a retrospective on how I found CVE-2025-39964 and how we developed th...

Keywords: af_alg vulnerability, vulnerability af_alg, vulnerable af_alg_tsgl, crypto af_alg, af_alg crypto, root af_alg, exploited kernel, af_alg linux, developed exploit, af_alg processes
Source: idnsec.com

Default Timeouts: Nginx, Apache, PHP, Cloudflare, AWS and More

A request that takes too long doesn’t fail at one place — it fails at whichever layer gives up first. The CDN, the load balancer, the web server, PHP and the database each have their own timeout, and...

Keywords: timeout nginx, php max_execution_time, max_execution_time php, timeout php, max_execution_time cloudflare, cloudflare timeout, servers max_execution_time, shorter timeout, timeout cloudflare, database timeout
Source: copahost.com

HTTP 413 Content Too Large (Request Entity Too Large): How to Fix It

You try to upload a file — a theme, a backup, a video, a big form — and instead of a success message you get 413 Request Entity Too Large. The request was fine. It was just bigger than something betwe...

Keywords: upload exceeds, upload limits, upload limit, request exceeds, upload_max_filesize 64m, withupload_max_filesize large, larger upload_max_filesize, large upload, large uploads, maximum upload
Source: copahost.com

Reverse-engineering the vintage Intel 8087's tangent algorithm: more than CORDIC

I hope you're not tired of the 8087, because I have another article about Intel's floating-point chip.1 In 1980, Intel introduced the 8087, making floating-point operations much faster in the IBM PC a...

Keywords: 8087 implemented, 8087 compute, 8087 calculations, 8086 microprocessor, intel 8087, implementation 8087, 8087 algorithm, faster 8087, 8086 speedup, 8087 uses
Source: righto.com

Understanding the Impact of LLM Watermarking on AI Agent Behavior

The Provenance Tax: Understanding the Impact of LLM Watermarking on AI Agent Behavior Recently, Anthropic announced that future Claude models would embed an invisible watermark in their output [1], [2...

Keywords: watermarking ai, agents watermark, watermarking provenance, behavior watermarking, evidence watermarking, watermarked model, models watermarking, feasible watermarking, watermarking introduced, argue watermarking
Source: lasso.security

Show HN: Reladraw – A diagram language where you decide where to place things

A text language for diagrams where you say where things go. Try it in your browser → — edit the source on the left, watch the layout re-solve on the right. Nothing to install. A diagram drawn by hand...

Keywords: understanding diagramming, diagramming, language diagrams, diagram reladraw, diagramming want, install diagram, reladraw diagram, especially diagram, diagrams, diagram format
Source: github.com

Leanpub Book LAUNCH 🚀 The Code Is Lying: Engineering with Coding Agents by Alen Joy

Leanpub Book LAUNCH 🚀 The Code Is Lying: Engineering with Coding Agents by Alen Joy When code gets cheap, understanding becomes expensive. AI can turn a description into working software in minutes. T...

Keywords: coding agents, agents want, automation goal, expensive ai, harder automate, ai turn, ai, leanpub launch, automation better, agentic ai
Source: leanpub.com

ML Papers Explained (Free PDF)

ML Papers Explained is an open GitHub collection created to make important machine learning papers and concepts easier to explore. Instead of presenting research papers as a long list, the repository...

Keywords: learning papers, ml papers, papers models, learning ml, papers explained, papers concepts, papers explore, papers understand, concepts papers, research papers
Source: clcoding.com

Mungert/OpenMath-Nemotron-7B-GGUF — public verification state: pending

Mungert/OpenMath-Nemotron-7B-GGUF Mungert/OpenMath-Nemotron-7B-GGUF is an open-weights language model for general under the cc-by-4.0 license (132.9 GB), added to Hugging Bay Jun 30, 2026. Open AI mod...

Keywords: openmath nemotron, download openmath, mungert openmath, openmath, q4_k_m openmath, nemotron 7b, download mungert, openmathreasoning, openmathreasoning popularity, openmathreasoning paper
Source: huggingbay.xyz

MM Change Slated For Linux 7.4 Yields +22904539.81% In One Metric, More Modest Wins In Others

MM Change Slated For Linux 7.4 Yields +22904539.81% In One Metric, More Modest Wins In Others Queued this week into the slab/for-next Git branch ahead of next month's Linux 7.4 merge window is a memor...

Keywords: benchmarks sheaves, sheaves cpu, benchmarks patch, barn benchmarks, barn_get performance, sheaf barn, allocator sheaves, barn sheaf, kernel memory, sheaves barn
Source: phoronix.com

Ollaya runs open decision models locally behind a Jev-compatible API

Ollaya runs open decision models locally behind a Jev-compatible API Maintainer Mert Cobanov packages open decision models behind a TypeSafe-compatible API, with speed and calibration claims that stil...

Keywords: software ollaya, ollaya compatible, ollaya engineering, ollaya deployments, ollaya says, ollaya matters, comparison ollaya, compatibility ollaya, matters ollaya, ollaya models
Source: runtimewire.com

Six System Design Problems, and the New Problem Each Fix Creates

Six System Design Problems, and the New Problem Each Fix Creates Every system starts the same way: a client talks to a server, and the server talks to a database. That shape carries most products furt...

Keywords: caching strategies, slower writes, cache replica, performance scalability, scale writes, heavy writes, lags writes, stale reads, replication lag, design cache
Source: devops-daily.com

Go concurrency distilled

Go concurrency distilled This mini-book provides a brief overview of many concurrency topics in Go. Each topic comes with interactive examples — feel free to experiment with them by changing the code...

Keywords: concurrency goroutines, concurrency goroutine, threads goroutines, goroutines threads, concurrency beginner, concurrency book, synchronizing goroutines, concurrency, handles concurrency, concurrency easiest
Source: antonz.org

Terraform State: Remove, Move and Migrate Resources, and Set Up a Remote Backend

Terraform State: Remove, Move and Migrate Resources, and Set Up a Remote Backend Terraform state is the file that maps every resource in your configuration to a real object somewhere: an instance ID,...

Keywords: terraform locking, terraform state, moved terraform_data, terraform manage, terraform_data resource, removed terraform_data, terraform forget, terraform lock, use_lockfile terraform, resource terraform_data
Source: devops-daily.com

optimum-intel-internal-testing/tiny-random-qwen3.5-moe-mtp — public verification state: pending

optimum-intel-internal-testing/tiny-random-qwen3.5-moe-mtp optimum-intel-internal-testing/tiny-random-qwen3.5-moe-mtp is an open-weights language model for general under the apache-2.0 license (29 MB)...

Keywords: weights language, optimum intel, open weights, moe mtp, mtp optimum, qwen3 moe, download optimum, intel internal, llm optimum, big optimum
Source: huggingbay.xyz

FlagRelease/Xing4.0-29B-A4B-BF16-zhenwu-FlagOS — public verification state: pending

FlagRelease/Xing4.0-29B-A4B-BF16-zhenwu-FlagOS FlagRelease/Xing4.0-29B-A4B-BF16-zhenwu-FlagOS is an open-weights language model for general under the apache-2.0 license (62.4 GB), added to Hugging Bay...

Keywords: download flagrelease, flagrelease xing4, flagrelease added, flagos files, flagos library, flagos flagrelease, flagos unknown, flagos open, files flagrelease, install huggingbay
Source: huggingbay.xyz

Josephgflowers/Phinance-Phi-4-mini-instruct-finance-v0.4-with-reasoning — public verification state: pending

Josephgflowers/Phinance-Phi-4-mini-instruct-finance-v0.4-with-reasoning Josephgflowers/Phinance-Phi-4-mini-instruct-finance-v0.4-with-reasoning is an open-weights language model for general under the...

Keywords: download josephgflowers, files josephgflowers, download hugging, metadata josephgflowers, install huggingbay, source download, downloads explicit, reasoning files, downloads, model josephgflowers
Source: huggingbay.xyz

HTTP 429 Too Many Requests: What It Means and How to Fix It

429 Too Many Requests means you’ve been rate-limited: you sent more requests in a short time than a server is willing to answer. Nothing is broken — the server is working, and it’s asking you to slow...

Keywords: 429 requests, requests limit, limit requests, requests constantly, limits requests, requests slow, requests burst, http 429, request rate, requests limit_req
Source: copahost.com

No errors, no warnings, no gods, no masters - HTML Purity is a Fetish

"The problem with defending the purity of the HyperText Markup Language is that HTML is about as pure as a cribhouse whore. The standard doesn't just borrow concepts from better languages; on occasion...

Keywords: html validity, valid html, html valid, language html, pure html, abused html, html pure, html best, hypertext markup, html write
Source: shkspr.mobi

AI Agents Just Rediscovered the Dual-Write Problem

This agentic case starts with a request no business owner wants, but still we should plan for: a customer asks an AI support agent for a refund. And the agent does exactly what it is told to do: it lo...

Keywords: agent refund, refund agent, agent recover, agent issues, agent continue, agent workflow, agent run, refund succeeds, agent tries, failure completes
Source: newsletter.systemdesignclassroom.com

Claude Opus 5.5 across every effort level | VulcanBench Frontier v4

September 26, 2026 · 115 runs · Code quality protocol v3.15 Claude Opus 5.5 across every effort level Across every effort level Claude Opus 5.5 ran every Frontier v4 task once at each of five effort l...

Keywords: effort levels, quality tasks, cost task, effort level, task effort, cost opus, score tasks, task level, code quality, high runtime
Source: vulcanbench.com

If we do not stop to help each other, what do we become?

If we do not stop to help each other, what do we become? Yesterday, I received this email as a response to You Can't Vibe Code Love. It's such a remarkable and powerful statement that I asked permissi...

Keywords: deployed zamboanga, community, family doing, help colleagues, come family, hope family, difficult interviewing, community really, fight job, leave important
Source: blog.codinghorror.com

Show HN: A Claude Code skill to analyze your chess games

Claude Code skills that turn one of your chess games into a post-mortem you can actually read: plain-language explanations of your mistakes, checked against Stockfish, and a narrated video of the whol...

Keywords: chess analysis, cp chess, chess postmortem, python chess, claude games, write chess, claude code, claude transcribed, lichess chess, chess
Source: github.com

One Month Without AI

Several months ago, I decided that AI contributions were no longer welcome in a FOSS project I am building and maintaining - LibreWeddingPlanner. It’s not that it got a lot of contributions with AI —...

Keywords: ai contributions, contributions ai, ai addiction, ai developers, ai work, work ai, ai foss, use ai, efforts ai, ai decreased
Source: blog.bustikiller.com

NetBSD Playing with disklabels

blog - git - desktop - contact 2026-09-25 Coming from DOS and Linux (and having largely ignored this on my OpenBSD systems -- yeah, shame on me), I'm not very familiar with the BSD disklabels. So let'...

Keywords: bsd disklabels, systems disklabels, netbsd disklabel, disklabel netbsd, disklabels usual, disklabels stored, sys disklabel, disklabels, disklabel metadata, tells disklabel
Source: movq.de

The JavaScript Pun: tagged template literal

Shakespeare’s chock-full of puns, no doubt as a sign of his mastery and love of words. Rarely can you swap two words to form a correct grammatically sentence, yet he played with words like a magician....

Keywords: definition pun, simplest puns, words puns, puns play, pun string, adjective witty, witty fool, overplayed pun, shakespeare chock, humorous effect
Source: shukla.io

Reading List 2026-09-26

Welcome to the reading list, a weekly roundup of news and links related to buildings, infrastructure, and industrial technology. This week we look at permitting reform prediction markets, ASML’s lack...

Keywords: manufactured homes, constructed homes, housing states, manufactured home, homes allowed, buildings infrastructure, homes oecd, allowed housingwire, housing act, process residential
Source: construction-physics.com

Is your Postgres migration safe or not safe?

◆migration.sql CHECKING ⌥⇧F⋯ bash+⌃` $ npx safe-not-safe check migration.sql parse libpg_query 17 (wasm) · initializing · 0 statements · 301 chars note Downloading and compiling the PostgreSQL WASM pa...

Keywords: libpg_query wasm, postgresql wasm, libpg_query 17, compiling libpg_query, libpg_query, parse libpg_query, checking bash, postgresql parser, compiling postgresql, bash npx
Source: safenotsafe.dev