TechPicks (by Tal Koren)

TechPicks 13/09/2026 Anthropic CEO outlines plan to slow AI development https://techcrunch.com/2026/09/12/anthropic-ceo-outlines-plan-to-pace-the-frontier/ Anthropic: What will our economic future look like? https://www.anthropic.com/institute/econ-scenarios Inside Claude's Android app: Finding the Internal Settings https://nima-ahmadi.github.io/posts/claude-android-internal-settings/ Introducing the Agents API https://openai.com/index/introducing-the-agents-api AI cybersecurity is a cat and mouse game https://stackoverflow.blog/2026/09/11/ai-cybersecurity-is-a-cat-and-mouse-game/ Shopify is moving from React Native back to Swift and Kotlin https://shopify.engineering/back-to-native The Physics of Database Speed: from 300 to 1M transactions per second https://www.youtube.com/watch?v=vOEL_pHFYK0 Commander Keen's Adaptive Tile Refresh https://fabiensanglard.net/ega/index.html If AI can do math, what's the point of mathematicians? https://www.scientificamerican.com/article/mathematicians-confront-the-ai-apocalypse/ Integrations and feeds - Discord integration - https://techpicks.app/discord Slack integration - https://techpicks.app/slack RSS Feed - https://techpicks.app/rss Follow me on X - https://x.com/dewbjorn

TechPicks 14/09/2026 Tal Koren - How I use a single .zshrc file across macOS and Windows (WSL2) https://talkoren.com/blog/single-zshrc-macos-windows-wsl2/ David Sacks: OpenAI and Anthropic Don't Need Regulations to Pace Frontier Models https://twitter.com/DavidSacks/status/2098973625252708460 Aligned to whom? https://hyperbo.la/w/aligned-to-whom/ Trump rejects Silicon Valley’s calls for AI slowdown https://finance.yahoo.com/technology/ai/articles/trump-rejects-silicon-valley-calls-165148820.html Apple opens the door to a nightmare world of always-listening tech https://this.weekinsecurity.com/watch-what-you-say-apple-opens-the-door-to-a-nightmare-world-of-always-listening-tech/ A better way of blocking macOS updates https://zoey-on-github.github.io/blog/updates.html Integrations and feeds - Discord integration - https://techpicks.app/discord Slack integration - https://techpicks.app/slack RSS Feed - https://techpicks.app/rss Follow me on X - https://x.com/dewbjorn

TechPicks 15/09/2026 Claude Fable 5.1 Solves the 373 year old Cyphral Distich https://www.vals.ai/blogs/fable-solves-cyphral-distich macOS 27: new Siri takes on AI productivity apps https://techcrunch.com/2026/09/14/macos-27-new-siri-takes-on-ai-productivity-apps/ When the fractional part of a float fixes your shader https://crocidb.com/post/when-the-fractional-part-of-a-float-fixes-your-shader/ The case against JPEG XL https://giannirosato.com/blog/post/case-against-jxl OpenAI Reveals There Was a Second Rogue AI Incident, Even Before Hugging Face: 'More' May Be Out There https://www.techtimes.co.uk/openai-agents-accessed-rubygems-internal-testing-1808674 Integrations and feeds - Discord integration - https://techpicks.app/discord Slack integration - https://techpicks.app/slack RSS Feed - https://techpicks.app/rss Follow me on X - https://x.com/dewbjorn

TechPicks 16/09/2026 Zuck says AI can't slow down, Dario at Anthropic says it should, and Anthropic is about to IPO at like a $2T valuation https://www.axios.com/2026/09/14/anthropic-ipo-safety-openai Hugging Face is billing OpenAI $100mn for hacking it https://thenextweb.com/news/hugging-face-delangue-openai-100m-compute-traces-demand Small Programming Tricks https://will-keleher.com/posts/small-programming-tricks-matter/ Cisco email security boxes can be rooted by... an email https://www.theregister.com/security/2026/09/15/cisco-email-security-boxes-can-be-rooted-by-an-email/5296604 New insights from Google's AI & Economy ATLAS https://blog.google/innovation-and-ai/technology/ai/ai-economy-atlas-september-2026/ A simple way to identify slow Postgres SQLs https://bookofrevenue.com/blog/6aa8d8b292d31a00012a8642/identify-slow-postgres-query Integrations and feeds - Discord integration - https://techpicks.app/discord Slack integration - https://techpicks.app/slack RSS Feed - https://techpicks.app/rss Follow me on X - https://x.com/dewbjorn

TechPicks 20/09/2026 - OpenJev - Local decision model in your browser - openjev.​com Introducing Astra for Law - openai.​com Researchers used Claude to hack OpenAI - arstechnica.​com Could AI really kill us all? Your questions, answered. - technologyreview.​com OpenAI caught its models leaving notes to successors to hide bad behavior - tech.​yahoo.​com Gemini Hacked Three Companies in First Known Breakout by Google’s AI - simonwillison.​net Claude Code revamps projects so you can work and pay in parallel - theregister.​com Open Weights Are Good. Open Source Is Better - opensource.​org Making a game for the GBA and PC from the same codebase - mattgreer.​dev Follow TechPicks - Website: This edition RSS: Feed X: @dewbjorn

Weekly recap Three links this week got a lot of interest, and they all ended up circling the same question from different angles - how much of the work are we handing over, and what do we get back? Top of the list by a decent margin is OpenJev - Local decision model in your browser, a browser demo that runs open local language models entirely on your GPU through WebGPU, using wllama and GGUF-quantized weights pulled from Hugging Face. What I like here is the actual experiment - it compares reading choice logits directly against generating a JSON distribution token by token, across Qwen3 0.6B, MiniCPM5 2B and Qwen3.5 4B, then reports balanced-accuracy numbers against a hosted proprietary model called Jev. All the inference and timing happens on the client side. The project also goes out of its way to say it's independent and not endorsed by TypeSafe, which, given how often that disclaimer gets quietly omitted elsewhere, is worth a nod. It's since been renamed to SemIf, so don't be confused if the name doesn't match. Then there's How we made claude.ai 3x faster in two weeks, where Anthropic ran a two-week performance sprint with an internal agent working autonomously in Slack threads (btw, have you added TechPicks to Slack yet?), landing over 3,000 changes across web, desktop, Claude Code and Claude Cowork. The interesting part, and I wish more people would take this away from the post, is that none of it works without measurable targets: instruction counts, layout shift telemetry, frame budgets, plus guardrails like automated code review, feature flags and CI ratchets. Give something a number to optimize against and it can optimize. The bug hunting is impressive too - an em-dash triggering a UTF-16 string slowdown (we all have gotten used to disliking em-dashes, amirite?). Thirteen measured user journeys got faster, and the ratchets are staying in place to stop the usual slow slide back. Lastly, AI Has No Wisdom and Neither Will You is a nice counterweight. Alexa

TechPicks 27/09/2026 - Revealing the details of how OpenAI agents hacked Hugging Face - swarmtraces.​org Ollaya – Ollama for open-source, Jev-style decision models - ollaya.​dev Analyzing Frontier Model Progress with My Favourite Game: Prince of Persia - blog.​priyan.​in Microservices are organizational debt disguised as architecture. - martinfowler.​com We still maintain a development tool first released in 1993. Here’s what 30+ years of backwards compatibility looks like - visualneo.​com How to keep enjoying programming in a world of LLMs - discourse.​haskell.​org Follow TechPicks - Website: This edition Add to Slack: techpicks.app/slack Submit a url: techpicks.app/submit X: @dewbjorn

TechPicks 29/09/2026 - Sonnet 5.5 - anthropic.​com Yes, no AI is now a feature - blog.​documentfoundation.​org How to Sync a Design System with Claude Design - nitayneeman.​com The problem is not the AI code, but nobody knows anything anymore - ssp.​sh A-ha, Take On Me - recreated in JavaScript - dittytoy.​net VoiceStudio - open-source voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages - github.​com Thieves steal Nvidia-branded trailers expecting valuable hardware, find 40,000 pounds of sand instead - techspot.​com Follow TechPicks - X: @techpicksapp Website: This edition Add to Slack: techpicks.app/slack Submit a url: techpicks.app/submit Saw a pick a friend needs? Invite them over: t.me/techpicks

TechPicks 30/09/2026 - OpenAI scraps GPT-6.1 Astra release over safety concerns - theverge.​com OpenAI apologizes after AI model breached Medicare and state systems - smh.​com.​au Dots: Always-on agents - openai.​com What Would A Serious AI Product Look Like? - blog.​glyph.​im Corral - Kill every command your agent starts - github.​com Beyond Light and Dark (mode) with CSS tokens - jlopes.​eu macOS Golden Gate Is a Buggy Mess - squareorbits.​com Follow TechPicks - X: @techpicksapp Website: This edition Add to Slack: techpicks.app/slack Submit a url: techpicks.app/submit Saw a pick a friend needs? Invite them over: t.me/techpicks

TechPicks 05/10/2026 - Pi - An AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI - github.​com Kolibri: A Sovereign Open-Weight Model - aleph-alpha.​com Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s - github.​com Why we are still arguing over CSS in 2026. - andreasmoller.​dk I Tested 11 HTTP Resilience Libraries - blog.​gaborkoos.​com Why don't more developers “use the platform”? - nolanlawson.​com Turn off Apple Intelligence on macOS 27 and get its disk space back - github.​com Car is a smartphone on wheels. Here's who's listening - automatictransmission.​khoury.​northeastern.​edu A browser-native classic Visual Basic VB6 IDE - wieslawsoltes.​github.​io Follow TechPicks - This edition | Share channel | Slack | X | Submit a url Support TechPicks: github.com/sponsors/dewbjorn

TechPicks 06/10/2026 - Claude Says - ohhfishal.​net OpenAI "rogue" agent activities found on Wikimedia projects - diff.​wikimedia.​org War on Peace: The Push Notification Strikes Back - godobject.​dev CSS does your tooltip positioning now - allthingssmitty.​com Web Search API - developers.​cloudflare.​com Follow TechPicks - This edition | Share channel | Slack | X | Submit a url Support TechPicks: github.com/sponsors/dewbjorn