Latest articles on AI, technology, and software development.
Enterprise buyers are starting to ask vendors to prove their binaries can be rebuilt bit-for-bit from source, and SLSA gives procurement teams a checkbox to demand it.
Nearly half of developers now use an AI reviewer on their pull requests. The time saved on reading code is being spent triaging what the AI flagged instead.
MCP crossed 97 million monthly SDK downloads and every major AI provider now ships native support. But transport, auth, and permission models still diverge per implementation — turning 'build once, use anywhere' into 'build once, test separately per client.'
Teams are shipping more AI-written code than ever while trusting it less than ever — and verifying it even less than that. The data on what happens when that gap meets a CI pipeline is not encouraging.
88% of AI coding agent pilots never reach production — and the real bottleneck isn't model intelligence, it's that monorepos built for humans are hostile to fixed-context AI agents.
38% of cloud-native organizations now run WebAssembly in production, up from 4% in 2022. The Component Model is quietly solving the cross-language plugin problem that containers never could.
As coding agents gain longer autonomy windows and broader filesystem access, isolated worktrees and sandboxed execution have shifted from a nice-to-have to required infrastructure. Here's what a real setup looks like and why the naive approach fails.
Cloudflare and Fastly independently arrived at the same architecture: Wasm instead of containers at the edge. Here's the technical reason why, and what it means for how you should ship code in 2026.
Startup latency, memory safety, and single-binary distribution are pulling CLI tool development away from Go and Python toward Rust in 2026.
With over 60% of newly committed code now AI-assisted, the constraint on shipping software has moved from typing speed to review capacity — and a new category of AI review tools is emerging to close the gap.
Models like Qwen3-Coder, Gemma 3, and Phi-4 now run on a single laptop or workstation GPU and handle routine coding work well enough that many developers default to local-first, reserving cloud APIs for the hardest problems.
AI code review has crossed from novelty to default practice at most engineering shops, but tool-level false positive rates of 5-15% mean teams are trading review speed for a new kind of review debt.