Daily AI changelog for engineers
While you're building, the stack is changing. New models, APIs, and tools ship daily. We watch so you don't have to. Free, every morning.
Recent reports (161)
See all →Qwen3.5-397B open weights land – 1M-context Plus priced $0.60/M input
Alibaba open-weighted Qwen3.5-397B-A17B under Apache 2.0; specs circulated as 397B total params with 17B active MoE plus native multimodal I/O; UI tab...
ByteDance Seed 2.0 prices $0.47 input, $2.37 output – 79-page benchmarks
ByteDance quietly rolled out the Seed 2.0 “frontier” family (Pro/Lite/Mini plus a dedicated Code variant) via Doubao; a 79‑page model card and screens...
MiniMax‑M2.5 releases 229B MoE open weights – $1/hr at 100 tok/s
MiniMax open-sourced MiniMax‑M2.5, positioning it as an “always‑on” long‑horizon agent backbone priced at $1/hour for ~100 tok/s; the model card cites...
GLM‑5 open weights debut at 744B – $1 input per 1M tokens
Z.ai launched GLM‑5 (“Pony Alpha”), a text‑only MoE flagship at 744B params trained on 28.5T tokens; it ships with 200K context and up to 128K output,...
We read 1000+ accounts. You get the summary.
A free daily report for engineers building with AI. The stack changes fast — this keeps you current.
Last week: 47 releases tracked · 12 breaking changes flagged · 3 pricing drops caught
Why engineers choose AI Primer
Built for builders who need signal, not noise
Agent tooling updates
Claude Code, Codex, Cursor releases — new configs, flags, and fixes that matter
Infrastructure signals
Compute announcements, rate limits, and scaling patterns from the front lines
Skills & MCP ecosystem
New integrations, skill syncs, and extension drops across agent harnesses
Workflow patterns
Context management, plan hygiene, and prompt discipline from power users