Prime Radiant

Building for a community: a code of conduct and a usage meter

Week of July 27, 2026 · Jul 27, 2026–Aug 2, 2026 · published Aug 2, 2026

No releases this week — but two brand-new repositories that both point outward. A community code of conduct signals that these projects expect outside contributors, and agentic-usage-meter, a macOS menu-bar app for tracking coding-agent subscription quotas, went public. Underneath the new repos, serf logged another heavy direct-push week and a research campaign in superpowers-autoresearch closed out with a measured, substantial cost cut.

The week in numbers

  • 14 repos active (2 featured, 12 also shipped)
  • 1,833 commits
  • 5 merged PRs
  • +370,575 / −48,783 lines
  • 6 commit identities
  • 0 tagged releases

prime-radiant-inc/code-of-conduct

The Prime Radiant Community Code of Conduct.

5 commits +161 −30

A small repo with an outsized signal: the Contributor Covenant (v3) adapted into a Prime Radiant community code of conduct, refined after review feedback. You don’t publish a code of conduct for a codebase only you touch — you publish it when you expect people to show up. With outside contributors already turning up in the commit logs, the timing tracks.

prime-radiant-inc/agentic-usage-meter

A macOS menu-bar meter for coding-agent subscription quotas.

139 commits +51,624 −6,917

Multi-provider quota tracking stood up in a single week: OpenCode account qualification and profile-lifecycle handling, SuperGrok credential refresh without reconnecting, and sample-data marking in account settings. It also arrived with real public-release infrastructure — signed Sparkle auto-update releases with fail-closed automation, confirmation-based updates, privacy-safe product screenshots, and public documentation. Built to be shipped, not just demoed.

Also shipped

prime-radiant-inc/serf

A non-interactive coding agent: give it a prompt and it reads, writes, runs commands, and searches code in a loop until the work is done, using native tool-calling across OpenAI, Anthropic, and Google models.

1,176 commits +158,943 −36,618

A second consecutive four-figure-commit week, nearly all direct push. The dominant visible thread was the web UI’s compact input footer — designed, constrained, documented, then implemented with simplified session-footer facts. Around it: mobile viewport and shell-chrome pinning, job-notification descriptions surfaced in notifications, task-mutation summaries showing task titles, and a refinement of how shell tool-calls are presented.

prime-radiant-inc/superpowers-autoresearch

The automated-research harness for running experiment campaigns against the superpowers methodology.

202 commits +94,251 −1,377

A heavy week that closed out a “queue campaign” measuring agent cost-pathologies — pre-registered batteries and verdicts across batching, plan-framed prompts, and at-scale queuing, landing on a headline result that a batching arm cut cost substantially while improving completion. A related cost-pathologies thread covered parser-scope limits, an unencrypted “honoring channel” design amendment, and scope-auditability notes.

prime-radiant-inc/teststrip

A macOS photo-culling app — catalog-first, non-destructive, AI-assisted. Early development; not ready for use.

94 commits +19,423 −2,086

Per-face report cards shipped end-to-end: a sharpness/light/facing/prominence analyzer, a staleness-aware FaceReportStore, traffic-light chip presentation, and burst-rail roll-up dots. A real dogfooding-found bug got fixed — serialized CoreImage face detection to break a live-only deadlock blocking a culling scenario — alongside a separate “blaze-through” thread: sliding-window burst prefetch and a warm-set planner for the cull loupe.

prime-radiant-inc/superpowers-evals

A behavioral eval lab (Quorum) that drives real coding-agent CLIs through a QA agent and grades them on workflow compliance against scenario criteria and deterministic post-checks.

2 commits +4,943 −1

Closed out a “codex-efficiency” measurement campaign with a batch of experiment write-ups, and bumped the pinned Codex CLI version. A small commit count carrying a large documentation payload.

obra/lace

A lightweight agentic coding environment.

4 commits 4 PRs +758 −21

Runtime-correctness fixes: stop mislabeling a notification as “a new message,” never dispatch an assistant prefill while draining a mid-turn inject, keep persona MCP servers alive across session resume, and forward container mounts so the mount-conflict scan sees the truth.

obra/winpepper

Windows-native local dictation: hold a hotkey, speak, release, and cleaned-up words appear in the focused app — Parakeet TDT v3 ASR with LlamaSharp cleanup, all local.

182 commits +39,369 −1,686

Another heavy week, mostly from outside contributor Dan Shapiro: model-aware cleanup-settings honesty (graying out unsupported controls when a raw-IO cleanup model is active), a CPU-pegged status-pill indicator, a multi-model ASR comparison harness (per-model resource capture, convergence detection, aggregated report), and mid-paste focus-change handling that halts and parks a pending click-to-paste.

prime-radiant-inc/stockyard

A coding-agent VM orchestrator: runs coding agents in isolated VMs — Firecracker micro-VMs on Linux with ZFS-based audit-trail snapshots, and Apple’s container tool on macOS.

3 commits +97 −13

Continued fail-closed task-destruction hardening: scoping the destroy-confirmation claim to the destroy command, and pinning the destroy-output substrings that automation consumes.

obra/dotfiles

Jesse’s personal dotfiles.

16 commits +977 −20

A curated Brewfile for one-command new-Mac bootstrap (including the clearance cask from Prime Radiant’s own Homebrew tap), a Bitwarden clipboard-paste walkthrough, and an Arq snapshot-cleanup utility.

prime-radiant-inc/homebrew-tap

A Homebrew tap for Prime Radiant tools, including formulae for llm-proxy and beeper-message-sync.

1 commit +18 −0

Added a cask (clearance).

obra/superpowers

An agentic skills framework and software-development methodology that works.

1 commit 1 PR +0 −6

A quiet week after the release: one merged PR removing the “We’re Hiring” section from the README.

prime-radiant-inc/smevals

A framework for running evals against small (and large) models.

1 commit +4 −1

Added badges to the README — small front-door housekeeping.