Hi there, this is your daily ☕️ Devshot.
In today's Devshot:
🚫 Reddit kills RSS feeds
⚡ AWS takes aim at TypeSafe's Jev model
☁️ Cloudflare launches Basin analytics suite
🦀 Ubuntu 26.10 beta embraces Rust
📌 Rust 1.100 drops 32-bit Windows MSVC
Plus: 🎁 7 more stories you might like, 🧰 6 tools, and 📚 5 papers.
🚫 Reddit kills RSS feeds LINK
- Reddit will shut off its RSS feeds on November 13 and close public API access in March 2027, cutting the main routes third-party tools, bots, and alerts have used for years to pull data.
- New public API requests end October 31, unregistered apps start losing access on January 12, 2027, and Reddit is pushing developers onto its Devvit platform with a $1 million migration fund paying $1,000 per eligible app that moves over.
- Over 14,000 apps have registered and bots like RemindMeBot have nearly finished migrating, but for RSS use outside communities you moderate Reddit says there is no replacement, so unmigrated tools go dark; migration applications were due November 30.
⚡ AWS takes aim at TypeSafe's Jev model LINK
- Amazon has released Strands Decider 2B, an open-source decision model that sits inside an agent workflow to judge whether a proposed action should proceed, directly challenging TypeSafe's hosted Jev model.
- Fine-tuned from a Qwen3.5-2B base, it strips the next-word prediction head and swaps in a small "pointer" component, just over a million parameters trained via a rank-16 LoRA, that scores supplied answer options in a single forward pass.
- Free under Apache 2.0 with published training data and scripts, it runs locally in tens of milliseconds, but AWS hasn't shown it beats Jev on JevBench (~72% accuracy versus Mapika's higher score) nor undercuts Jev's unusually low hosted price once your own hardware is counted.
☁️ Cloudflare launches Basin analytics suite LINK
- Cloudflare made its serverless analytics suite, now renamed Basin, generally available today, bundling event ingestion, Apache Iceberg storage, and distributed SQL querying atop R2 Object Storage with no egress charges.
- Basin splits into three services: Pipelines ingests from Workers, HTTP, or Logpush and now handles up to 3GB/s per stream, Catalog manages Iceberg metadata and compaction, and SQL runs over 190 functions across tables.
- Usage-based pricing bills only when Basin ingests, processes, or queries data, with no hourly or infrastructure fees, and you can read or write via any Iceberg-compatible engine including PyIceberg, DuckDB, Snowflake, and Apache Spark.
🦀 Ubuntu 26.10 beta embraces Rust LINK
- The Ubuntu 26.10 "Stonking Stingray" beta completes Canonical's switch to Rust-based coreutils, moving the last GNU Coreutils holdouts, cp, mv, and rm, to the uutils implementation by default.
- Ubuntu can now claim 100 percent Rust coreutils out of the box, and the beta also pulls the Rust-based Sequoia PGP into the main archive, with sq and sqv positioned as eventual replacements for gpg and gpgv.
- Everyday file operations should behave exactly as before while gaining memory-safe internals; the older GNU tools stay available in the main repository for now, with the final release expected on October 15.
📌 Rust 1.100 drops 32-bit Windows MSVC LINK
- Rust 1.100.0 will stop shipping host tools like the compiler for 32-bit Windows, demoting i686-pc-windows-msvc to Tier 1 without host tools and i686-pc-windows-gnu to Tier 2 without host tools.
- Prebuilt standard library builds will still be distributed and i686-pc-windows-msvc remains under CI testing, but the toolchains themselves can no longer be installed or run on 32-bit Windows hosts after this release.
- Developers targeting 32-bit Windows must now cross-compile from a supported host such as a 64-bit Windows toolchain, a shift driven by i686 compiler binaries crashing and the GNU C++ toolchain hitting OOMs during LLVM builds.
Other news you might like
- Introducing Clef: our open-source decision models, and new RL fine-tuning platform LINK
- The Pulse: RoR creator sparks new “death of coding by hand” debate LINK
- Pi Durable LINK
- Announcing Cloudflare K2: serverless event streams LINK
- Read Restrictions and Catalog Labels: Unifying governance across engines and catalogs LINK
- SequenceHash: multihashing for the rest of us LINK
- Why do OpenAI's GPT-2 weights beat mine? Part five: data quality LINK
🧰 Trending tools
Jev: makes fast, structured AI-driven decisions for software automation, helping developers build reliable logic into workflows without manually writing rules LINK
Switch: open-source workspace connecting people and AI agents across Slack, Teams, Discord, or Mattermost, preserving context during handoffs without migrating your existing tools. LINK
Meridian: open-source, locally-run activity tracker that logs work with context, showing where your time goes and which blockers stall projects. LINK
Monospace from Directus: provides an instant REST and GraphQL API plus a no-code data app for any SQL database, giving apps and teams a governed layer. LINK
iFixAi: audits AI agents with 250 inspections across 69 misalignment categories, flagging failures and giving engineers evidence to investigate and fix them. LINK
Tines 3B: builds and runs team automations, apps, and AI agents in one secure platform, keeping workflows unified and easy to manage LINK
📚 Trending papers & reports
Quantum compiler trust gets a practical playbook comparing the main ways to catch bugs when software that translates quantum programs changes, warning that today's test results can't be ranked against each other. LINK
Executable spec checks give scientific coding assistants ready-made tests of requirements instead of written instructions, lifting successful bug fixes and cutting the assistant's output by ~31% on physics tasks, though gains vary by task. LINK
Smart training schedules let AI agents improve faster by automatically picking which practice scenarios to drill next based on recurring failures, boosting task success rates by roughly 4 to 8 points over fixed training orders. LINK
AI agent performance depends on pairing the right model with the right surrounding software, with rankings flipping across tasks, as one model led by ~8 points in one setup but trailed by ~30 in another. LINK
Trustworthy data and AI operations for transportation and logistics get a comprehensive playbook covering the tools, practices, and reliability safeguards needed to run dependable machine-learning systems in real-world transport settings, serving operators, researchers, and policymakers. LINK
See you tomorrow for a new dose of ☕️ Devshot!