Blog

JevBench scores Jev-class decision models on four axes: accuracy above chance, calibration, speed and cost per 1,000 decisions. Its v1.4.1 board puts Jev first at 63.3. A day later, a post argued that no benchmark can make these probabilities trustworthy on your data. Both are right, and this post covers what to measure yourself.

• system-one-models
• jev
• benchmarks
• calibration

Since 2.1.277, Claude Code reads AGENTS.md only when no CLAUDE.md or CLAUDE.local.md sits in the working directory or above it. Until 2.1.281, a remote feature flag also skipped the file without a warning for anyone with telemetry off. Here is the precedence table, how to check what loaded, and which other agents read the same file.

• claude-code
• agents-md
• coding-agents
• context-engineering

A harness is everything in a coding agent except the model. Two September 2026 studies measured which parts of it matter: context management when the window is tight, planning mostly as a cost saver, and the harness as a whole far more for your bill than for your pass rate.

• coding-agents
• harnesses
• claude-code
• context-engineering

A week after launch, Jev had a trusted monitor, two computer-use agents, three self-hosted clones, a benchmark and a chatbot nobody should use. Here is the stack slot each project fills, and which of their numbers anyone has shown.

• system-one-models
• jev
• ai-agents
• benchmarks

Compaction is the step where a coding agent summarises its own history to keep working. Here is what it keeps, what it drops, what OpenAI found hiding in its summaries in September 2026, and the moves people use to compact less.

• coding-agents
• context-engineering
• claude-code
• compaction

Within a week of Jev's launch, open replications of it were everywhere. They are four different kinds of project with incomparable latency claims. Here is what each one replaces, and what you give up by leaving the hosted API.

• system-one-models
• jev
• open-source
• local-ai

TypeSafe launched Jev on 15 September 2026: a model that returns typed decisions instead of text. Here is what a System One model is, what is proven versus claimed, and where it sits in a stack.

• system-one-models
• jev
• llsm

The first edition of our AI coding tool switch series starts with a very small September 2026 sample. Cursor and Claude Code lead, while co-use is more common than switching.

• ai-coding-tools
• coding-agents
• cursor
• claude-code

Stackness is on Product Hunt today. What it is, what shipped in the four weeks since we opened, and the three kinds of feedback that would actually help.

• product-hunt
• stacks
• moves

Stackness launches on Product Hunt tomorrow, so we did what everyone does before guests arrive: tidied up. Public trending pages, category hubs, move cards for your README, tile quick look and a nicer blog.

• changelog
• product-hunt
• trends

Tools are what you install. Moves are how you actually work: the habits, rituals and workflow tricks that make a stack yours. What a move is, and how to write one worth stealing.

• moves
• workflows
• getting-started

Your stack page is now a bento grid you arrange yourself - subspaces, five tile sizes, drag-to-group clusters, all edited in place. Plus bookmarks everywhere, one History tab, and full MCP parity so agents can arrange your profile too.

• changelog
• profile
• stacks

A new picker fills your stack from what people run together. Tap the tools you use, watch related ones show up next to them, publish the lot as one update.

• stack-builder
• stacks
• getting-started

August is supposed to be the quiet month. Instead we got a compiler changing languages, a protocol getting audited in public, and Hacker News once again deciding that Postgres is the answer to every question. Here is what actually moved in developer stacks this month - not the product launches, but the shifts that change what you would put in a new project today.

• monthly-roundup
• trends

The official launch of the platform

• getting-started
• stacks
• moves