|
|
Hi there, this is your daily ☕️ DevShot.
|
|
|
In today's DevShot:
|
|
🐧 Epic starts building Fortnite for Linux ⚡ AWS Lambda now serves web apps directly 🎨 Figma launches Specs to guide AI design 🚀 Relace Apply 3 hits 10k tokens/second 🐬 Cloudflare fixes rare Hyperdrive MySQL bug Plus: 🎁 8 more stories you might like, 🛠️ 5 tutorials, 🧰 7 tools, and 📚 5 papers.
|
|
Discover laptops with built-in intelligence.Shift your workforce into hyper-productivity mode with Dell Pro laptops powered by Intel® Core™ Ultra with Intel vPro®. Shop on Dell Premier.
|
🐧 Epic starts building Fortnite for Linux
LINK
|
- Epic is laying the groundwork to bring Fortnite to Linux, starting with a rebuilt Easy Anti-Cheat system and a new vibe-coded Epic Games Launcher that already ships in preview for the platform alongside Windows and Mac.
- The new EAC Linux v2 uses eBPF to detect cheating and protect game processes, but demands signed kernels and modules, TPM 2.0, Secure Boot, measured boot, IOMMU support, and kernel lockdown to satisfy security concerns.
- Epic will test EAC Linux v2 with Fortnite first before opening it to other developers, with Ubuntu, Fedora, Debian, openSUSE, and Universal Blue already recognized, though there is no release date and the Steam Deck currently lacks required features.
|
⚡ AWS Lambda now serves web apps directly
LINK
|
- AWS Lambda now runs standard HTTP web apps directly through new Web Functions, letting you deploy a port-listening Express, Hono, Fastify, Next.js, or any Node.js server with one command and no event-handler rewrites, adapters, or framework lock-in.
- Each execution environment handles up to 128 concurrent requests so traffic bursts share warm environments, and billing meters only active CPU time, so waiting on a database query or API call consumes little, cutting costs for I/O-heavy workloads like server-side rendering and streaming inference endpoints.
- Available today on Node.js 24 Arm64 in five regions, Web Functions auto-replicate across regions with latency-based routing through one global URL, support canary deploys by splitting endpoint traffic between immutable revisions via a single API call, and work as CloudFront private origins using OAC.
|
🎨 Figma launches Specs to guide AI design
LINK
|
- Figma launched Specs, a Markdown editor that lets teams document their design language and brand rules so Figma's own and third-party AI agents can consult them and produce more consistent output.
- Specs documents live as standalone files or on the canvas, and can reference canvas elements and embed FigJam diagrams, code layers, and designs, centralizing design libraries, code libraries, and guidelines for agents to read while working.
- The tool aims to steer non-deterministic AI toward predictable results, arriving alongside Figma's wider on-canvas agent rollout; Crisan said teams now generate 17 times more lines of code than before while shipping only 1.3 times as many releases.
|
🚀 Relace Apply 3 hits 10k tokens/second
LINK
|
- Relace open-sourced its fine-tuning methods alongside Relace Apply 3, a specialized code-merge model that runs at over 10k tokens per second while holding state-of-the-art accuracy on applying LLM-generated diffs back into existing files.
- Rather than have a frontier model regenerate a whole file, over 100 seconds and at least $0.18 for a thousand-line edit with Claude 4.5 Sonnet, a small 3-8B model infers diff intent, handling pathological edits a fixed string-replace or UDiff algorithm would botch.
- Relace trained via LoRA on ~145k production-distribution examples filtered by a Claude 4 Sonnet judge tuned to a ~1% false-positive rate, then quantized the weights to FP8 to exploit newer Nvidia GPU cores for throughput with no measurable accuracy loss.
|
🐬 Cloudflare fixes rare Hyperdrive MySQL bug
LINK
|
- Cloudflare has moved Hyperdrive's MySQL support to general availability after squashing a rare caching bug that returned empty query responses, bringing its reliability in line with the connection pooler's existing Postgres handling.
- The bug surfaced only when a Hyperdrive Client and Endpoint shared a data center cache and one read the other's cached bytes, which were encoded with mismatched MySQL capabilities, like the mysql2 driver lacking CLIENT_DEPRECATE_EOF, causing early EOFs or hangs roughly 0.001% of the time.
- The fix caches a canonical response encoded with Hyperdrive's own capabilities regardless of which node stores it, pushing Client connection success to 99.99%; you can now connect to any MySQL-compatible database via the dashboard or wrangler.
|
|
State of Product reveals what AI still hasn’t solved80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.
📚 Tutorials & Advice
|
> Building a fancy scrollspy using modern CSS: Build a scroll-position-aware navigation highlight entirely in CSS using new scroll-marker properties, dropping JavaScript while keeping plain anchor links working in unsupported browsers.
|
|
|
🔗 Other News & Articles
-
SRE book second edition
LINK
-
Headless observability & autonomous operations
LINK
-
Ubuntu 26.10 released with Linux 7.3 + GNOME 51, Rust Coreutils migration complete
LINK
-
From 2,500 requests to 24: Rebuilding star history without the stargazers
LINK
-
Apple contributes SSAF driver to LLVM Clang for source code transformations
LINK
-
Harbor + Docker sandboxes: Run safe, reproducible agent evals
LINK
-
Uber’s AI infra, from the inside
LINK
-
CedarDB wants one database to handle both transactions and analytics, as AI agents raise the stakes
LINK
|
|
📄 Papers & Resources
|
> Open visual-coding models turn text or screenshots into working charts, webpages, and animations, scoring the generated code on whether it runs, is well built, and looks right once displayed.
LINK
|
|
> Coding assistants produce far less varied problem-solving approaches than human programmers, capturing only ~31% of human diversity on their own, meaning heavy reliance on them may narrow the range of strategies developers learn.
LINK
|
|
> Automatic code cleanup gets a new test showing top commercial AI can clear up to 90% of style and error warnings but breaks working code a third of the time, exposing a reliability gap.
LINK
|
|
> Repo-specialist coding models get trained to master one codebase deeply rather than many shallowly, letting compact 4-billion-parameter versions match or beat rivals up to 8x bigger while running privately on limited hardware.
LINK
|
|
> Framework migration gets a tough new test that checks whether code rewritten from PyTorch to JAX actually behaves identically, and the best coding agent scored only ~0.68 against a ~0.98 perfect benchmark.
LINK
|
|
|
|
🧰 Latest Tools
| ElevenLabs: turns any script into studio-quality speech in seconds, with 5,000+ voices across 70+ languages. Free to start. GENERATE FOR FREE |
|
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: open-source AI coding agent for JetBrains IDEs that lets you build and iterate using 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 sends push notifications when sessions need input, letting you dictate prompts and attach files remotely.
LINK
|
|
akta.pro: provides a private company data API with event tracking, news, and entity resolution, helping finance and GTM teams source deals faster
LINK
|
|
dif.sh: open source feature flags stored as markdown files beside your code, reviewed in PRs, with context your coding agent reads automatically.
LINK
|
|
|
|
💬 How did you find today's edition?
We read every reply — just reply to this email and let us know how we can improve!
|
|
|