โ† All digests

Dev Digest โ€” April 16, 2026

๐Ÿ”ฅ HOT RELEASES

Claude Opus 4.7 Anthropic's most capable Opus model yet โ€” better at long-running tasks, more precise instruction following, self-verifying outputs. Same pricing as 4.6. Claude Code 2.1.111 ships alongside with new xhigh effort level and persisted debug logs. ๐Ÿ”— X: https://x.com/claudeai/status/2044785261393977612 ๐Ÿ”— Anthropic: https://www.anthropic.com/news/claude-opus-4-7

Qwen3.6-35B-A3B (Open Source) Alibaba drops a sparse MoE model: 35B total params, only 3B active. Apache 2.0 license. Agentic coding on par with models 10x its size, strong multimodal reasoning. Runs efficiently on consumer hardware. ๐Ÿ”— X: https://x.com/Alibaba_Qwen/status/2044768734234243427 ๐Ÿ”— HuggingFace: https://huggingface.co/Qwen/Qwen3.6-35B-A3B

VS Code 1.116 โ€” Agents Go Stable Big release: the Agents app is now in Stable, powered by CLI as the agentic engine. Copilot Chat is built-in (no longer a separate extension). Agent debug logs persist via JSONL. New carousel UI for tool approvals. Groundwork for multi-agent parallel execution. ๐Ÿ”— X: https://x.com/code/status/2044555141039190157 ๐Ÿ”— X (deep dive): https://x.com/OrenMe/status/2044637007108850040

Rust 1.95.0 Released today with if-let guards in match statements, cfg_select! macro, expanded match capabilities, and new stable APIs. ๐Ÿ”— X: https://x.com/rustaceans_rs/status/2044726405137596610 ๐Ÿ”— Blog: https://blog.rust-lang.org/2026/04/16/Rust-1.95.0/

Tailscale-rs โ€” Rust Library for Embedded Networking Tailscale releases a Rust library (with Python/Elixir/C bindings) for embedding secure Tailscale networking directly into apps. Safer, lighter alternative to Go's tsnet. ๐Ÿ”— X: https://x.com/rustaceans_rs/status/2044775327595925875 ๐Ÿ”— Blog: https://tailscale.com/blog/tailscale-rs-rust-tsnet-library-preview ๐Ÿ’ป GitHub: https://github.com/tailscale/tailscale-rs


๐Ÿงช INTERESTING REPOS

Strix โ€” AI-Powered Pentesting Open source AI security scanner that finds and fixes app vulnerabilities. Now with context-aware pentesting that understands your architecture and business logic. ๐Ÿ”— X: https://x.com/7h3h4ckv157/status/2044641154734682497 ๐Ÿ’ป GitHub: https://github.com/usestrix/strix

OpenFang โ€” Autonomous Agent OS Open-source operating system for autonomous agents. ๐Ÿ”— X: https://x.com/tom_doerr/status/2044576783484252290 ๐Ÿ’ป GitHub: https://github.com/RightNow-AI/openfang

Twenty โ€” Open Source CRM (Notion-like) CRM designed to feel like Notion. Clean UI, extensible. ๐Ÿ”— X: https://x.com/tom_doerr/status/2044553514119012749 ๐Ÿ’ป GitHub: https://github.com/twentyhq/twenty

PureMac โ€” macOS App Cleaner (No Telemetry) Open-source macOS app cleaner. Zero telemetry. ๐Ÿ”— X: https://x.com/tom_doerr/status/2044615561787605279 ๐Ÿ’ป GitHub: https://github.com/momenbasel/PureMac

Sidebar โ€” AI Live Stream Commentary Real-time AI commentary by different personas on YouTube live-streams. Open source, run locally. ๐Ÿ”— X: https://x.com/marckohlbrugge/status/2044547840219684868 ๐Ÿ’ป GitHub: https://github.com/marckohlbrugge/sidebar

VoiceBox โ€” Local Voice Synthesis Studio Open-source local voice synthesis. No cloud dependency. ๐Ÿ”— X: https://x.com/tom_doerr/status/2044568810707955863 ๐Ÿ’ป GitHub: https://github.com/jamiepine/voicebox

FinGPT โ€” Financial LLMs Open-source large language models fine-tuned for finance. ๐Ÿ”— X: https://x.com/tom_doerr/status/2044638948119413242 ๐Ÿ’ป GitHub: https://github.com/AI4Finance-Foundation/FinGPT

BenchJack โ€” AI Benchmark Exploit Toolkit Tool that finds and exploits scoring vulnerabilities in AI benchmarks (SWE-bench, WebArena, OSWorld). Red-teaming tool for researchers building tamper-proof evals. ๐Ÿ”— X: https://x.com/lihanc02/status/2044480184288743692


๐ŸŽฅ WORTH WATCHING

Triton GPU Programming From Scratch Full tutorial on writing custom CUDA-style kernels in Python with Triton. Covers memory layout, strides, tiling, loop unrolling. Make models train 2-10x faster. ๐Ÿ”— YouTube: https://youtu.be/LMU0LrX1f88 ๐Ÿ”— X: https://x.com/VukRosic99/status/2044650862195376357

Claude Code Desktop Deep Dive Chris Coffee tests the new Claude Code desktop setup โ€” multiple sessions, integrated terminal, task management, preview. "Feels like a real workspace now." ๐Ÿ”— X: https://x.com/ChrisCoffee/status/2044501400013463736


๐Ÿ’ก TECHNIQUES & IDEAS

Simon Willison: Clone Repos into Claude.ai Tell Claude to "clone x/y from GitHub" and it can answer questions about the repo or use code snippets to build new artifacts. Handy for exploring unfamiliar codebases. ๐Ÿ”— X: https://x.com/simonw/status/2044574333247557916 ๐Ÿ”— Blog: https://simonwillison.net/2026/Apr/16/datasette-io-preview/

TypeScript Transformer v3 Rewrite Ruben rewrote TypeScript Transformer โ€” now feels like a bridge between PHP and TypeScript instead of a one-way code generator. ๐Ÿ”— X: https://x.com/freekmurze/status/2044747398459908420

The Copy-Validate-Improve-Differentiate Flow @alexcooldev's dev workflow: copy success โ†’ validate fast โ†’ make it better โ†’ make it different. ๐Ÿ”— X: https://x.com/alexcooldev/status/2044588733522215124


๐Ÿ”ฎ EMERGING TRENDS

Cal.com Goes Closed Source โ€” "AI Security" Excuse Cal.com closed its core codebase citing AI-powered vulnerability discovery. Community split: some understand, many call it security-through-obscurity theater. Community fork lives at cal.diy. Could signal a wave of OSS projects reconsidering open code. ๐Ÿ”— X: https://x.com/initjean/status/2044436903726907807 ๐Ÿ’ป Community fork: https://github.com/calcom/cal.diy

The Model Wars Intensify Opus 4.7 + Qwen3.6 + GPT-5.5 rumors all in one day. Polymarket running bets on release timing. OpenAI reportedly dropping something soon after Anthropic (again). Bolt.new says Opus 4.7 scored highest ever on complex builds.

AI Agents Eating Security Multiple signals: Strix for automated pentesting, BenchJack for benchmark exploitation, SpankRAT using Rust for stealth. The "agents that audit code" paradigm is accelerating โ€” both offense and defense.

Anthropic Requires Government ID New Claude users now need government ID + photo verification. No regulatory mandate โ€” Anthropic chose this. Controversial. ๐Ÿ”— X: https://x.com/Cointelegraph/status/2044511136096149649


Full digest: https://digest.99.cooking