Hi there, this is your daily ☕️ Devshot.
In today's Devshot:
🔒 Apple tightens macOS access controls
🗄️ Doom runs in SQL in 5,900 lines
🦅 Google adds server-side support to Swift
🤝 Supabase acquires Turso
🛠️ Chrome 141 previews DevTools MCP server
Plus: 🎁 6 more stories you might like, 🧰 6 tools, and 📚 5 papers.
🔒 Apple tightens macOS access controls LINK
- Apple will add new restrictions on macOS's Full Disk Access permission, the catch-all setting that lets an app read or write nearly any file on a Mac's drive, citing abuse by developers and the growing risks from autonomous AI agents.
- Originally meant for backup apps, the permission sidesteps Apple's privacy controls and exposes files, mail, messages, contacts, and browsing history; updaters from Adobe, Logitech, and Microsoft OneDrive request it, as do most AI agents on the Mac.
- Apple hasn't detailed the controls or a timeline, but the company says granting such access will require very explicit user action going forward, likely starting with stronger warning language spelling out exactly what an app can see.
🗄️ Doom runs in SQL in 5,900 lines LINK
- Lukas Vogel's SQLDoom project runs the 1993 game's logic inside CedarDB, a Postgres-compatible database, in 5,900 lines of SQL versus the original C source's 9,000, with a Python frontend handling only graphics, sound, and input.
- Entity updates that needed "for" or "while" loops in C collapse into a single parallel "UPDATE... WHERE" statement, while the 1,300-line renderer is one query spanning 89 tables, with BSP traversal reduced to a "SELECT... ORDER BY" sorting walls front-to-back.
- Multiplayer came out cleaner than the original, since syncing interdependent state is what databases do, snapshots, authentication, and access control are free, and a game tick is just a transaction, though floor and ceiling flood-fill renderers didn't map well to SQL.
🦅 Google adds server-side support to Swift LINK
- Google has released Google Cloud API Client Libraries for Swift, backing the language as a server-side option and letting developers build GCP infrastructure with native async/await APIs and compile-time data-race safety.
- The libraries target Swift 6.2+ and lean on SwiftNIO's asynchronous event loops, HTTP/2 multiplexing, gRPC transport, and strict concurrency flags, catching data races in the editor before a binary compiles or reaches production.
- Google pitches the SDK for microservices built with Swift frameworks like Hummingbird or Vapor, though the article notes server-side Swift adoption has been slow and its TIOBE ranking slipped from 9th in January 2020 to 18th.
🤝 Supabase acquires Turso LINK
- Supabase has agreed to acquire Turso, the startup behind a SQLite-based database built for AI agents, to extend its own AI data capabilities alongside a $150 million funding round.
- Turso inherits SQLite's small hardware footprint, so agents can spin up instances almost instantly, and it adds embedding support plus per-page encryption that aren't available in stock SQLite.
- Supabase plans to fold the technology into Supabase Compute, its new hosted-sandbox service for long-running agents, letting them launch lightweight Turso instances for simple tasks that don't need a full relational database.
🛠️ Chrome 141 previews DevTools MCP server LINK
- Chrome 141 ships a preview of the Chrome DevTools MCP server, letting AI coding assistants tap DevTools debugging and performance insights directly inside the browser.
- Coding agents previously couldn't observe how their generated code behaved at runtime; the MCP server lets them debug web pages live in Chrome, improving accuracy when identifying and fixing issues.
- This release also adds a Debug with AI button on the Network dependency tree insight that opens the Gemini-backed AI assistance panel, where you can now export or copy the conversation.
Other news you might like
- Introducing Web Search API via AI Gateway LINK
- How Booking.com cut Node.js costs by 38% with Watt LINK
- Aspire 13.6 Adds Persistent Dashboard Telemetry and First-Party Java and Rust Hosting LINK
- Introducing Cloudflare Traces: follow requests through our entire platform LINK
- The Agent Said It Was Done. The Database Disagreed. LINK
- The original PlayStation 2 security chip has been reverse engineered LINK
🧰 Trending tools
Die With Me: a Mac app that tracks Claude and Codex usage, compares it with friends, and opens a chatroom when your allowance runs out. LINK
Kilo Code for JetBrains: an open-source AI coding agent for JetBrains IDEs, letting you build and iterate with flexible model choices across projects. LINK
Mastra Factory: a TypeScript framework for building AI agents with staged workflows, memory, streaming, evals, tracing, and a Studio UI for testing. LINK
Kilo Code: open-source mobile AI coding agent that pushes notifications when sessions need input, letting you dictate prompts and attach files remotely LINK
Floot MCP: links Floot with Claude and ChatGPT to build full-stack apps from chat, handling database, logins, hosting, emails, and app-store publishing LINK
HarnessRouter: open-source unified API that connects Codex, Claude Code, Hermes, and other agent harnesses through one interface, so teams skip managing separate agent backends. LINK
📚 Trending papers & reports
Trading-strategy code checks reveal that even a top model builds the wrong logic ~28% of the time, running clean backtests while silently executing a strategy the trader never described. LINK
Combinatorial test quality is best judged by one metric, value combination coverage, which reliably predicts which test suites catch real software bugs, giving teams a trustworthy yardstick across nearly 300,000 test suites studied. LINK
VeriPy lets Python developers write correctness guarantees as ordinary code comments and automatically checks that updates still work for existing users, easing the maintenance burden of keeping formal proofs aligned with running code. LINK
Adversarial testing for coding agents generates fresh tests after each attempt an AI makes, catching behavior it quietly skipped and lowering failure rates versus writing all the tests once upfront. LINK
Bug-hunting shortcuts speed up finding which code change broke large software like Linux by searching recent commits first, cutting search steps ~26% on average since half of bugs hide in the newest ~0.7% of history. LINK
See you tomorrow for a new dose of ☕️ Devshot!