# 18 — Latest-versions refresh (2026-07-05 sweep) (https://jackin.tailrocks.com/reference/research/token-optimization-tools/18-latest-versions-refresh/) # 18 — Latest-versions refresh (2026-07-05 sweep) [#18--latest-versions-refresh-2026-07-05-sweep] This page records the 2026-07-05 latest-versions verification sweep across every tool in the hub: GitHub/registry API reads, release-note review, and targeted raw-file fetches for each tool covered in chapters 01–17 and 19–25, plus a new-entrants search over the 2026-06/07 window. The dated chapters [13 — Reuse surfaces and deltas (2026-07-03 source sweep)](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/) (2026-07-03 source sweep) and [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/) (2026-07-03 verification pass) stay frozen as-written; this chapter is the new dated layer on top, and where a delta invalidates a claim in chapters 01–16 or 19–25 the correction is registered here rather than edited in place. All sources in this chapter were accessed 2026-07-05; version/star/date facts are registry- and API-verified (T1-grade metadata), vendor claims quoted from READMEs and release notes remain T4 unless said otherwise. Timing note: three tools (caveman, headroom, sem) shipped releases on 2026-07-03 itself — the same day as the last sweep's clones — so several "current as of 07-03" statements were stale within hours of being written. ## TL;DR [#tldr] * **caveman shipped v1.9.1 on 2026-07-03** ("65%, honestly"), hours after the last sweep recorded it frozen at v1.9.0 — and formally **retired the \~75% headline** the hub spends a full section debunking, replacing it project-wide with a flat "65% average output-token reduction (range 22–87%), output tokens only" plus a new `docs/HONEST-NUMBERS.md` that concedes when caveman *costs* tokens. The "frozen flagship, stars pumping" framing in chapters [01 — Caveman: design teardown](/reference/research/token-optimization-tools/01-caveman-design/) and [14 — The caveman ecosystem: flagship verified, cavemem and caveman-code deep-dived](/reference/research/token-optimization-tools/14-caveman-ecosystem/) is stale; the 75%-vs-65% debunk now describes a claim the project no longer makes ([release](https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1)). * **cavemem is no longer unmaintained.** Revived 2026-07-03: issues mass-triaged 44 → 4, npm packaging bug fixed, opt-in web-enrichment MCP tool added, in-repo version bumped to 0.3.0 (npm still 0.2.1). [14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)'s "dead since 2026-05-17 / 44 unanswered issues" is now false; the measurement-based "refuted as shipped" critique stands until re-measured — the 07-03 commits are enrichment/Windows/packaging, not compression-engine changes ([commits](https://github.com/JuliusBrussee/cavemem/commits/main)). * **codedb's version narrative in [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/) is inverted**: v0.2.5827 (native Windows support, 2026-06-23) was already the latest GitHub release at the 07-03 sweep, and page 15's explicit "current is 0.2.5826, not 5827" correction is false against GitHub releases; the "history squashed to one commit; no changes since 2026-06-23" clause is also wrong (main shows 100+ commits of history; repo pushed 2026-07-04). Mechanism evidence holds — main HEAD is still the exact audited commit `c12e1e2` — and 0.2.5828 is in active preparation on a release branch. * **sem released v0.17.0 and v0.18.0 within \~19 hours of the last sweep** and **retired both hero claims** (2.3× accuracy, 75% fewer tokens) for paired-run SWE-bench Verified A/B speed numbers with token parity and unchanged solve rate stated plainly. Chapter [25 — sem and mq: entity-level retrieval and structural markdown queries](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)'s `=0.16.2` pin is two breaking 0.x minors behind and its T4 marketing caveat now describes retired marketing. * **headroom shipped v0.29.0 and v0.30.0 both on 2026-07-03** — `--lossless` graduated from Unreleased to a tagged release, and a new `--force-kompress-all` opt-in slightly softens the "ML being demoted" framing (gated, with an explicit force switch, not retired). **RTK** is unchanged at v0.43.0 but chapter [03 — RTK: design teardown](/reference/research/token-optimization-tools/03-rtk-design/)'s header table is still pre-0.43 and the `uv run` transparent prefix listed in [13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/) as an incoming 0.44 feature was reverted on develop. **lean-ctx** has v3.9.0 fully cut in-repo (AI Gateway team mode, grammar-addon dylibs, agents-only hook default) but not yet tagged or published — the largest single release since the sweeps began, imminent. * **Where nothing moved**: fff (last push 2026-07-02; `=0.9.6` pin current; nushell still pins it; 0.9.7 nightlies signal an imminent bump), mq (one weekly patch behind: v0.6.4), the observability set (minor drift; token-optimizer restructured its savings claims, token-dashboard now \~2.5 months dormant), and the thinking-token blind spot the hub flags as a jackin❯ gap **still holds** — no surveyed tool added a `count_tokens`-based thinking estimator. The new-tools sweep found exactly one genuine arrival (Recall) plus watch-list-grade entrants; **pxpipe** joins the comparison set with its own chapter, [17 — pxpipe: design teardown](/reference/research/token-optimization-tools/17-pxpipe-design/). ## Summary table [#summary-table] Recorded = the snapshot the hub's most-recent dated layer carries (chapter and date noted). Latest, pushed\_at, and stars/watchers are 2026-07-05 API reads. | Tool | Recorded snapshot | Latest (release date) | pushed\_at | Stars / watchers | Verdict in one phrase | | --------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------- | ---------- | ---------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | | caveman | v1.9.0, "frozen since 2026-06-12" ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/), 07-03) | v1.9.1 (2026-07-03) | 2026-07-03 | 83,839 / 194 | No longer frozen; \~75% claim retired for flat 65% | | cavemem | "dead since 2026-05-17" ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | 0.3.0 in-repo; npm still 0.2.1 (2026-07-03 revival) | 2026-07-03 | — (family repo) | Revived; unmaintained claim now false | | caveman-code | v0.65.2, active ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | v0.65.2 unchanged (npm 2026-05-18) | 2026-06-21 | 703 / 3 | Drifting toward dormant, no flip yet | | cavekit | 1,068★ ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | v4.1.0 tag (2026-06-18; GitHub release object still v3.1.0) | 2026-06-18 | 1,074 / 10 | Tag-only releases since April | | cavegemma | 73★, frozen ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | unversioned, no commits since 2026-05-17 | frozen | 79 / 0 | Still frozen; watch-don't-adopt stands | | headroom | v0.28.0 ([13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/), 07-03) | v0.30.0 (2026-07-03; v0.29.0 same day) | 2026-07-03 | 56,509 / 176 | Two releases post-sweep; all verdicts hold | | RTK | v0.43.0 ([13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/)/15) | v0.43.0 (2026-06-28); develop at 0.44-rc.308 | 2026-07-04 | 68,484 / 166 | Stable unchanged; [03 — RTK: design teardown](/reference/research/token-optimization-tools/03-rtk-design/) header stale, one 0.44-rc feature reverted | | lean-ctx | v3.8.18, "no commits since" ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | v3.8.18 (2026-07-01); v3.9.0 cut in-repo, unreleased | 2026-07-04 | 3,085 / 21 | Largest release since sweeps began is imminent | | codedb | "0.2.5826, not 5827" ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | v0.2.5827 (2026-06-23); 0.2.5828 branch active | 2026-07-04 | 1,341 / 9 | Version correction inverted; main frozen at audited commit | | sem | sem-core 0.16.2 ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/), 07-03) | v0.18.0 / sem-core 0.18.0 (2026-07-03) | 2026-07-04 | 3,054 / 6 | Two breaking minors; hero claims retired | | fff | v0.9.6, `=0.9.6` pin (engine spec + [16 — Global comparison](/reference/research/token-optimization-tools/16-global-comparison/)) | v0.9.6 (2026-06-21); 0.9.7 nightlies publishing | 2026-07-02 | 9,524 / 20 | No material change; 0.9.7 imminent | | mq | v0.6.3 ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)) | v0.6.4 (2026-07-03) | 2026-07-04 | 955 / 2 | One weekly patch behind; nothing recorded is false | | token-optimizer | "v5", 1,341★ ([23 — Token observability and session visualization](/reference/research/token-optimization-tools/23-token-observability-and-visualization/), 06-15) | v5.11.35 (2026-07-03) | 2026-07-03 | 1,542 / 8 | Savings claims restructured; blind-spot gap holds | | token-dashboard | 584★, live dashboard ([23 — Token observability and session visualization](/reference/research/token-optimization-tools/23-token-observability-and-visualization/)) | no releases; last push 2026-04-20 | 2026-04-20 | 619 / 11 | Dormant \~2.5 months; function unchanged | | claude-usage | 1,826★ ([23 — Token observability and session visualization](/reference/research/token-optimization-tools/23-token-observability-and-visualization/)) | v1.5.4 (2026-07-01) | active | 1,952 / 16 | Actively maintained, no material change | | pxpipe | — (new to the hub) | v0.8.0 (npm `pxpipe-proxy`, 2026-07-04) | 2026-07-04 | 1,728 / 9 | New entrant — full dossier in [17 — pxpipe: design teardown](/reference/research/token-optimization-tools/17-pxpipe-design/) | ## caveman family — deltas [#caveman-family--deltas] Three load-bearing deltas, all post-dating the 2026-07-03 sweep by hours. Confirmations that need no section: issue #484 (over-shrinking PR/issue titles) still open — not fixed in v1.9.1, though the release exempts build files from `/caveman-compress` (#600, a sibling fix); MIT licenses unchanged; no new caveman-family repos under JuliusBrussee since the 07-03 sweep (newest repos are Loop-Factory and skills, both 2026-06-08, not family members — negative result, [repos list](https://api.github.com/users/JuliusBrussee/repos?sort=created)). | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | -------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------ | | Flagship release freeze | "No release since v1.9.0 (2026-06-12)"; frozen flagship with stars pumping ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/) TL;DR, [01 — Caveman: design teardown](/reference/research/token-optimization-tools/01-caveman-design/) facts table) | v1.9.1 released 2026-07-03 — 50 changes: hook/installer hardening (JSONC string-aware merge, spawn ENOENT = failure, EXDEV fallback), NL-trigger misfire fixes, one-shot modes no longer stick, Hermes Agent support, per-agent cavecrew model overrides, slash commands now ship, SECURITY.md/no-telemetry statement | [https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1](https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1) | load-bearing | | Headline savings claim | README headline "cuts \~75% of output tokens"; hub repeatedly frames and debunks the pooled \~75% ([01 — Caveman: design teardown](/reference/research/token-optimization-tools/01-caveman-design/)) | v1.9.1 retired \~75% and the \~50–65% range project-wide (#234): headline is now a flat "65% average output-token reduction vs default verbose replies (range 22–87%), output tokens only" across README, manifests, skill descriptions, GitHub Pages; repo description now says "cuts 65% of tokens" | [https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1](https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1) | load-bearing | | cavemem maintenance | "Unmaintained since 2026-05-17; 44 open issues, zero maintainer replies"; "the entire project lived one month" ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | Revived 2026-07-03: last push 2026-07-03T14:47Z, open issues 44 → 4 (mass triage incl. #54 npm missing `dist/index.js`, #56/#57 Windows hooks → PowerShell 7), new opt-in web-enrichment MCP tool with SSRF/ReDoS hardening + metadata scrubbing (#55), IDE capture-vs-query capability matrix (#58); in-repo bump to 0.3.0 (npm still 0.2.1, 0.3.0 not yet published) | [https://github.com/JuliusBrussee/cavemem/commits/main](https://github.com/JuliusBrussee/cavemem/commits/main) | load-bearing | | `/caveman-stats` mechanism | "A local tally" computed by hook reading session JSONL ([01 — Caveman: design teardown](/reference/research/token-optimization-tools/01-caveman-design/)) | v1.9.1 fixes per-mode token attribution (#601), relabels the output-reduction line, adds a subscription limit-headroom meter, and ships `docs/HONEST-NUMBERS.md` documenting when caveman saves vs costs tokens and advising turn-off when net-negative | [https://github.com/JuliusBrussee/caveman/blob/v1.9.1/docs/HONEST-NUMBERS.md](https://github.com/JuliusBrussee/caveman/blob/v1.9.1/docs/HONEST-NUMBERS.md) | minor | | Adoption stats | 74,446★ / 166 watchers (2026-06-18); \~81.2k implied by 07-03 | 83,839★ / 194 watchers / 4,673 forks / 355 open issues+PRs; star-pump signal continues but the "zero commits" half of the signal no longer holds | [https://api.github.com/repos/JuliusBrussee/caveman](https://api.github.com/repos/JuliusBrussee/caveman) | minor | | caveman-shrink | Real and guarded, \~27k npm downloads/month ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | npm still 0.1.0 (2026-05-01), 23,282 downloads last month; v1.9.1 notes the published tarball was missing `spawn-options.js` it imported (#597, now shipped) — the npm package was broken as published until this landed repo-side | [https://api.npmjs.org/downloads/point/last-month/caveman-shrink](https://api.npmjs.org/downloads/point/last-month/caveman-shrink) | minor | | caveman-code | Active, v0.65.2, 671★ / 3 watchers, \~20.7k npm dl/mo ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | npm unchanged at 0.65.2 (still the only published version); last repo push 2026-06-21; 703★ / 3 watchers / 83 forks; 19,201 dl/mo — two weeks without a push, "active" drifting toward dormant but no flip yet | [https://api.github.com/repos/JuliusBrussee/caveman-code](https://api.github.com/repos/JuliusBrussee/caveman-code) | minor | | cavekit | 1,068★ ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | 1,074★ / 10 watchers; tags now include v4.0.0/v4.0.1/v4.1.0 (v4.1.0 commit 2026-06-18, "full loop" PR #33) though the latest GitHub *release object* is still v3.1.0 — tag-only releases since April; last push 2026-06-18 | [https://api.github.com/repos/JuliusBrussee/cavekit/tags](https://api.github.com/repos/JuliusBrussee/cavekit/tags) | minor | | cavegemma | 73★, LoRA Gemma fine-tune, watch-don't-adopt ([14 — The caveman ecosystem](/reference/research/token-optimization-tools/14-caveman-ecosystem/)) | 79★ / 0 watchers, no license file, zero commits since 2026-05-17, no tags/releases; description now pins target as Gemma 4 31B | [https://api.github.com/repos/JuliusBrussee/cavegemma](https://api.github.com/repos/JuliusBrussee/cavegemma) | none | Fix-worthy detail: v1.9.1's release notes confirm that several defects the hub treated as steady-state (broken caveman-shrink tarball, NL-trigger misfires, one-shot modes sticking) are now fixed. The 65% number caveman now claims matches the shape of the hub's own per-task-mean correction; the dossier's deeper corrections — the 58.5% Claude-tokenizer replication and the \~17%-of-dollars output-bucket ceiling — remain the load-bearing critique and are unaffected. ## headroom — deltas [#headroom--deltas] Core recorded architecture facts all hold: org move to `headroomlabs-ai` complete at repo level (badges still `chopratejas`, as [13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/) already notes), Apache-2.0 unchanged, Phase H (Python-proxy retirement) **not** executed (`headroom/proxy/server.py` still present at 219,879 bytes; `REALIGNMENT/10-phase-H-python-retirement.md` still lists the 3-PR deletion as pending), no crates.io publication (`headroom-core` does not exist; the bare crate name `headroom` is squatted by an unrelated audio-loudness tool v2.1.1 — the same squat pattern the hub recorded for rtk), and `kompress-v2-base` still on the founder's personal HF account (lastModified 2026-06-10, 1,761 downloads, ungated). The README-vs-PyPI headline inconsistency (H-K5: "60–95%" vs "50-90%") stands as written. No new independent benchmarks surfaced; the "47.5% was self-report" correction from [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/) stands unchallenged. | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | ----------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------ | | Latest version | v0.28.0 ([13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/), 07-03 sweep); v0.26.0 in [02 — Headroom: design teardown](/reference/research/token-optimization-tools/02-headroom-design/); v0.25.0 in [21 — Headroom and the context-compression layer](/reference/research/token-optimization-tools/21-headroom-and-context-compression/) | v0.30.0 (2026-07-03T19:22Z); v0.29.0 earlier the same day (06:30Z) — two releases landed on/after the last sweep's clone; PyPI `headroom-ai` 0.30.0, 195 releases total | [https://github.com/headroomlabs-ai/headroom/releases/tag/v0.30.0](https://github.com/headroomlabs-ai/headroom/releases/tag/v0.30.0) ; [https://pypi.org/pypi/headroom-ai/json](https://pypi.org/pypi/headroom-ai/json) | minor | | `--lossless` proxy mode | "New `--lossless` proxy mode in Unreleased" ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | Shipped in v0.29.0 (PR #1721, no-CCR mode with format-native compaction); v0.30.0 extends it with lossless-excluded compaction for grep/log/json wired into new workload personas (PR #1762) | [https://github.com/headroomlabs-ai/headroom/releases/tag/v0.29.0](https://github.com/headroomlabs-ai/headroom/releases/tag/v0.29.0) | minor | | ML demotion | kompress-v2 gated not replaced; TextCrusher opt-in default-off ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | Still gated, but v0.29.0 adds `--force-kompress-all` (route ALL content through kompress-v2-base) plus a fail-open guard when kompress saturation would exhaust pre-upstream budget (#1430) — upstream keeps the ML path alive as an explicit opt-in extreme rather than retiring it | [https://github.com/headroomlabs-ai/headroom/releases/tag/v0.29.0](https://github.com/headroomlabs-ai/headroom/releases/tag/v0.29.0) | minor | | Adoption stats | 33,359★ / 111 watchers ([02 — Headroom: design teardown](/reference/research/token-optimization-tools/02-headroom-design/), 06-18); 28,199★ ([21 — Headroom and the context-compression layer](/reference/research/token-optimization-tools/21-headroom-and-context-compression/), 06-15) | 56,509★ / 176 subscribers / 4,128 forks / 531 open issues; the watcher:star skew still supports the hub's PR-inflation read | [https://api.github.com/repos/headroomlabs-ai/headroom](https://api.github.com/repos/headroomlabs-ai/headroom) | minor | | Language split | Python \~78% / Rust \~17.3% / TS \~2.5% | Python \~80% / Rust \~15.4% / TS \~2.6% — Python share grew slightly; Phase H confirmed un-executed | GitHub languages API + REALIGNMENT/10-phase-H-python-retirement.md | none | New mechanisms in the two releases worth a line each: sk-ant-oat OAuth token-prefix auth fix (#1672); AVX2 gating for ONNX embedding + Magika backends to avoid SIGILL on old x86 (#1765/#1162); cache key now includes system/tools/sampling (#1473); unit-router path rejects lossy unmarked tool output (#1479); adaptive Otsu KEEP/DROP relevance threshold (#1726); npm `headroom-ai` clarified as TS-SDK-only, CLI pip-only (#1585); agent-savings workload personas for coding + general (#1732); Claude 5 family pricing (#1767). Post-0.30 main continues daily (opencode provider-config injection fix #1716, HEAD `e8151f0` 2026-07-03T22:20Z). Nothing in either release contradicts the hub's verdicts. ## RTK — deltas [#rtk--deltas] The 2026-07-03 sweep state holds almost entirely: v0.43.0 remains latest stable (0.44 unshipped; develop RC train at `dev-0.44.0-rc.308`, 2026-07-03), exactly 63 `.toml` files under `src/filters/` re-confirmed via the git tree API, RTK Cloud still waitlist-only (20 "waitlist" mentions on the homepage, no Cloud section in the develop README), issue #1155 (hook auto-allow permission bypass) still open, and the develop-branch `rtk/Cargo.toml` / `rtk/CHANGELOG.md` version drift persists — both still say 0.42.4, and the v0.43.0 release is absent from the in-repo changelog. Apache-2.0, `rtk-ai` org, default branch `develop`, and the 14-agent-integrations README claim all confirmed. | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | ----------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------ | | [03 — RTK: design teardown](/reference/research/token-optimization-tools/03-rtk-design/) header table | "Latest seen v0.42.4 (212 release tags)"; adoption 63,608★ / 146 watchers (2026-06-18) | Latest release v0.43.0 (2026-06-28); 68,484★ / 166 watchers / 4,232 forks / 1,485 open issues — page 03 is one stable release behind and internally inconsistent with pages 13/15/16, which record v0.43.0 | [https://api.github.com/repos/rtk-ai/rtk](https://api.github.com/repos/rtk-ai/rtk) ; [https://api.github.com/repos/rtk-ai/rtk/releases/latest](https://api.github.com/repos/rtk-ai/rtk/releases/latest) | load-bearing | | `uv run` transparent prefix | Listed in [13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/) as an incoming 0.44-rc feature | Removed on develop 2026-07-01: commit `96f9422` "fix(uv): remove uv run from transparent prefixes to fix rewrite conflicts" (PR #2768) — reverted after rewrite conflicts; should not be listed as an incoming 0.44 mechanism | [https://github.com/rtk-ai/rtk/commit/96f9422](https://github.com/rtk-ai/rtk/commit/96f9422) | minor | | Star:watcher ratio | 146 watchers, \~435:1 (chs. 03/15/24); 68,018★ / 1,464 issues ([13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/), 07-03) | 166 watchers, \~412:1; 68,484★ / 1,485 open issues — argument unchanged in substance, numbers drifted | [https://api.github.com/repos/rtk-ai/rtk](https://api.github.com/repos/rtk-ai/rtk) | minor | New develop mechanisms since the sweep clone `b52e02b`: TOML filters now matched **inside the PreToolUse hook itself** via a match-only RegexSet (PR #2748, merged 2026-07-03) — the declarative 63-filter corpus now participates in auto-rewrite command matching, not only direct rtk invocation; trust-before-load hardening extended (rtk trust reuses the init opt-in prompt, project- vs global-scoped labels, untrusted filters skipped silently on the command path but failing loudly when non-interactive, `rtk gain` flags untrusted custom filters); `git stash show` compressed to a compact diffstat (PR #2801); UTF-8 char-boundary panic fixes closing a CJK/emoji crash class (PR #2789 + `extract_json_object` byte-offset fix). Note an internal doc inconsistency this sweep could not resolve from the API alone: [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/) says 77 commands / 62 modules / 63 filters / 86 rules while [13 — Reuse surfaces and deltas](/reference/research/token-optimization-tools/13-reuse-surfaces-and-deltas/) says 42 native modules + 64 TOML filters + 85 rewrite rules — only the 63-filter figure was re-verifiable remotely; the rest need a clone audit to reconcile. ## lean-ctx — deltas [#lean-ctx--deltas] Docs are accurate on today's load-bearing facts: latest **released** version is v3.8.18 on both GitHub and crates.io, Apache-2.0, `yvgude/lean-ctx`, very active (\~30 commits in the last 48 h). The recorded tokbench R1 numbers are verbatim-confirmed in the live README ("pilot, N=1 per arm — replication runs in progress"); tokbench itself last pushed 2026-06-15 with no R2. The time-sensitive item: **v3.9.0 is fully cut in-repo** (CHANGELOG section dated 2026-07-04 plus a version-bump commit) but not yet tagged, released, or published — and it is the largest single release since the sweeps began. | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | -------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------ | | Release state | Clone `a80350da` == v3.8.18, "no commits since" ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)); [04 — lean-ctx: design teardown](/reference/research/token-optimization-tools/04-leanctx-design/) still says "Latest seen v3.8.9" | v3.8.18 still the latest *published* release, but \~30 commits landed 2026-07-03/04 including the v3.9.0 version bump and a full `[3.9.0] — 2026-07-04` CHANGELOG section; "no commits since" is now false, and [04 — lean-ctx: design teardown](/reference/research/token-optimization-tools/04-leanctx-design/)'s v3.8.9 is stale even by [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)'s own correction | [https://github.com/yvgude/lean-ctx/releases/latest](https://github.com/yvgude/lean-ctx/releases/latest) ; [https://crates.io/api/v1/crates/lean-ctx](https://crates.io/api/v1/crates/lean-ctx) | minor | | Commercial shape | Local engine free forever; paid tiers add only cloud sync/shared index/audit log — no gateway product | v3.9.0 ships **AI Gateway team mode** (`lean-ctx gateway serve/init/keys/evidence/gdpr`): multi-provider reverse proxy (Anthropic/OpenAI/Gemini/Ollama/custom), per-person SHA-256-hashed bearer keys, Postgres usage metering, signed org-policy gates (model-ceiling 403, budget-cap 429, per-person RPM), Ed25519 evidence exports, GDPR export/delete, Blake3 pseudonymization — compiled into the default binary; changelog asserts the local-free invariant intact. Materially changes the product-shape description once it ships | [https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md) | minor | | Footprint / grammars | 27 static tree-sitter grammars compiled in; 75.9 MiB measured from the v3.8.18 artifact; upstream "\~17 MB" claim unfounded ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | v3.9.0 adds **grammar addons**: long-tail grammars as SHA-256-pinned per-platform runtime dylibs (hash checked on every dlopen + ABI check, zero-config fetch with egress opt-out, world-writable-path refusal, offline regex fallback); registry ships empty. Phase 2 — moving any of the 27 built-ins out of the binary — is explicitly telemetry-gated and **not done**, with new per-extension `grammar_usage.json` telemetry as its groundwork. The 75.9 MiB figure stands; no v3.9.0 artifact exists to re-measure | [https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md) | minor | | Shell hook default | Hook compresses git/cargo/npm/docker/kubectl at the Bash boundary, always active once installed | Post-3.9.0 unreleased change (GH #699): default activation is now **agents-only** — hook aliases auto-activate only when an agent session is detected (6 env markers incl. CLAUDECODE, CURSOR\_AGENT); plain human terminals get verbatim commands; `shell_activation="always"` restores old behavior | [https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md) | minor | | Corruption bug class | Failure modes listed: map over-compression, stale graph, proxy prose rewrite, daemon fragility | Two new confirmed-in-changelog corruption incidents worth adding to the hub's failure-mode inventory: (1) GH #688 — sed/awk range-prints of source fell into the generic terse dictionary and had code identifiers word-substituted (function→fn, return→ret, bare else dropped); fixed by classifying sed/awk/gawk/mawk/nawk as file viewers. (2) GL #1158 — marked\_block healing matched markers by substring and live-wiped \~75 lines of the project's own AGENTS.md when the marker was quoted in prose; fixed to whole-line matching. Plus GH #685: `max_ram_percent` unenforced during index builds — 75 GB OOM-kill-respawn cycles on 1M+-file roots; now batched builds + RSS-floor eviction | [https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md) | minor | | Adoption stats | 2,800★ / 19 watchers / 278 forks / 13 issues ([04 — lean-ctx: design teardown](/reference/research/token-optimization-tools/04-leanctx-design/), 06-20) | 3,085★ / 21 watchers / 294 forks / 34 open issues+PRs | [https://api.github.com/repos/yvgude/lean-ctx](https://api.github.com/repos/yvgude/lean-ctx) | minor | Other v3.9.0/unreleased mechanisms relevant to the engine spec: `/v1/compress` wire-compatibility with LiteLLM ≥ v1.92 pre\_call prompt-compression guardrail; `ctx_read repo=` multi-repo parity (read-only by design); MCP roots/list retry re-arming; lost-update fix routing all knowledge writers through a single locked mutator; recovery-first refusal doctrine extended to heredoc-to-interpreter refusals; PowerShell signatures (#679). ## codedb — deltas [#codedb--deltas] The mechanism and install-surface evidence in [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/) holds by proxy: default branch main is **frozen at the exact audited commit `c12e1e2`**, and the main..release/0.2.5828 diff touches no install.sh/telemetry/hook files. What is wrong is the version narrative, in both directions. | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ | ------------ | | Latest release | [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/): "current is 0.2.5826, not 5827"; clone `c12e1e2` == v0.2.5826 | GitHub releases/latest is **v0.2.5827** "native Windows support", published 2026-06-23 (tag `4fb0be7` on the release lineage, not on main) — the "not 5827" correction is false against GitHub releases; a `release/0.2.5828` branch is actively merging PRs through 2026-07-04 ("Windows warm daemon, 2x codedb\_context, OSTree roots") | [https://github.com/justrach/codedb/releases/tag/v0.2.5827](https://github.com/justrach/codedb/releases/tag/v0.2.5827) | load-bearing | | Maintenance freshness | "No changes since 2026-06-23" ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | Repo pushed 2026-07-04; heavy activity on release/0.2.5828 and perf/\* branches (PRs #641–#654). Main-anchored claims (install hook, telemetry upload, README) remain accurate since main is unchanged | [https://api.github.com/repos/justrach/codedb](https://api.github.com/repos/justrach/codedb) | minor | | Provenance | "History squashed to one commit" ([15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | Live main at the same HEAD shows full multi-commit history (100+ commits via the commits API); since a commit sha encodes its ancestry, `c12e1e2` cannot both be a squashed single commit and carry this history — the squashed-history claim does not match the live repo | [https://api.github.com/repos/justrach/codedb/commits?sha=main](https://api.github.com/repos/justrach/codedb/commits?sha=main) | minor | | Platform claim | Local daemon/MCP with macOS/Linux runtime characteristics; no Windows mention | v0.2.5827 ships a native x86\_64-windows binary (POSIX/Windows platform shim: SRWLOCK, QueryPerformanceCounter, CreateFileMapping mmap shim; warm-daemon proxy disabled on Windows, CLI runs cold-direct); 0.2.5828 adds Windows warm-daemon IPC (#641) | [https://github.com/justrach/codedb/releases/tag/v0.2.5827](https://github.com/justrach/codedb/releases/tag/v0.2.5827) | minor | | Install-surface opt-out | No project opt-out ("the anti-pattern spec", [15 — Verification pass and the extended matrix](/reference/research/token-optimization-tools/15-verification-pass-and-extended-matrix/)) | On 2026-07-04 upstream merged `.codedbrc disable = true` + installer `--no-register` (commit `9ed2e74`) then **reverted it the same day** (`37aa83e`) on release/0.2.5828 — the recorded no-opt-out surface still stands, but upstream is visibly experimenting with softening it. If it lands in 0.2.5828, the anti-pattern framing needs a caveat | [https://github.com/justrach/codedb/commits/release/0.2.5828](https://github.com/justrach/codedb/commits/release/0.2.5828) | none | | npm launcher lag | "npx launcher" listed as an agent interface ([19 — Code-intelligence tools](/reference/research/token-optimization-tools/19-code-intelligence-tools/)) | npm `codedeebee` dist-tag latest is 0.2.5823 (2026-05-29), four releases behind, even though the v0.2.5827 release notes claim `npx codedeebee@0.2.5827` resolves the right binary — that npm version does not exist; a manual npm-publish workflow was added on main but apparently not run | [https://registry.npmjs.org/codedeebee](https://registry.npmjs.org/codedeebee) | minor | Confirmed unchanged: BSD-3-Clause license; default-on telemetry upload to codedb.codegraff.com, global PreToolUse hook blocking 10 shell commands, \~5.16k-token MCP schema, 1.7 GB RSS figures — all by proxy via the frozen main HEAD, not re-measured live. Curiosity: an anomalous tag `v0.2.58181` points at the same sha as v0.2.5819 — a typo tag, harmless but it pollutes tag listings. ## sem — deltas [#sem--deltas] Structural picture unchanged: dual MIT/Apache-2.0, Ataraxy-Labs org, daily solo-founder velocity, telemetry default-on with `SEM_NO_TELEMETRY`/`DO_NOT_TRACK` opt-out, cloud tier present (now framed as scale/team/CI upgrade). The cli/mcp crates.io publishing wedge persists exactly as recorded — sem-cli and sem-mcp frozen at 0.14.1 (2026-06-23) while sem-core ships 0.18.0; sem-cloud-client still unpublished (404). The attention-ledger mechanism description stands (no ledger changes in 0.17/0.18 changelog sections). | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | ------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------ | | sem-core version | 0.16.2 latest; reuse verdict pins `=0.16.2`; "\~50 releases to v0.16.2 in 5 months" ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)) | sem-core 0.18.0 published 2026-07-03T20:23Z (v0.17.0 and v0.18.0 both tagged 2026-07-03, hours after the sweep). The `=0.16.2` pin is two breaking 0.x minors behind; 0.16.0 already broke `ContextResult` struct-literal constructors (new public `omitted` field) | [https://crates.io/api/v1/crates/sem-core](https://crates.io/api/v1/crates/sem-core) ; [https://github.com/Ataraxy-Labs/sem/releases/tag/v0.18.0](https://github.com/Ataraxy-Labs/sem/releases/tag/v0.18.0) | load-bearing | | Hero claims | "2.3× more accurate agents" circular n=3 self-benchmark; retired-era "75% fewer tokens" framing — "cite the mechanism, not the number (T4)" ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)) | Both hero claims retired. Benchmarks page rebuilt on July-2026 paired-run SWE-bench Verified A/B data (grep+read agent vs sem agent, hidden-test graded): 50–65% faster code understanding, verify loop 2.90 s → 0.59 s/iteration when call-graph edges resolve (bimodal, 1.2× floor disclosed), token parity stated plainly, unchanged solve rate disclosed. The hub's caveat now describes marketing that no longer exists; the replacement claims are more honest and differently shaped (speed, not accuracy/tokens) | [https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md) | load-bearing | | Host-write surface | `sem setup` rewires `git diff` itself (the HN-controversial host write — not adopted) ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)) | Unreleased on main: `sem setup` now additionally installs two hooks into `~/.claude/settings.json` (SessionStart spawns `sem mcp --resident` detached; UserPromptSubmit runs context injection), with backup/idempotency and `sem unsetup`. The strip-list item's host-write surface grew from a git alias to Claude Code settings edits — strengthens, not weakens, the existing exclusion | [https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md) | minor | | Distribution | CLI, 6 MCP tools, GitHub Action, Claude skill, optional PreToolUse guard, sem setup ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)) | Now also published to the official MCP registry as `io.github.Ataraxy-Labs/sem`, discoverable from registry-browsing clients (VS Code, Cursor, Claude Code) — a distribution channel absent from the recorded interception list | [https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md) | minor | | Adoption | 3,032★ / 6 watchers / 93 forks ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)) | 3,054★ / 6 watchers / 93 forks; open issues 4; pushed 2026-07-04 | [https://api.github.com/repos/Ataraxy-Labs/sem](https://api.github.com/repos/Ataraxy-Labs/sem) | minor | New mechanisms relevant to the engine spec's read-stub/prefetch layer: `sem orient --pack ` — turn-zero packed briefings from raw issue/task text (code-ish term extraction, body-term-convergence ranking; IDF weighting, `
` stripping, name-echo slot in 0.18.0); unique-method-name call edges for Python/Ruby (a method name with exactly one repo-wide definition resolves to an edge, ambiguity yields none); `Parent::child` qualifiers accepted everywhere `Parent.child` works; `sem impact --tests` lexical fallback labeled as weaker evidence; git reftable ref-storage support. ## mq — deltas [#mq--deltas] [25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/) was researched 2026-07-03 — the same day v0.6.4 shipped — so it is exactly one weekly patch release behind. Nothing recorded is false: MIT license, solo maintenance (v0.6.4 changelog entirely harehare + dependabot; releases 6/14, 6/21, 6/24, 7/03), no format gating at 0.6.4 (csv, hcl-rs, toml, yaml-rust2, quick-xml, toon-format all non-optional in `mq-lang`'s Cargo.toml), the non-optional nushell `mq-markdown` dependency (now 84,131 downloads), and mq-db still a 0.1.x prototype (0.1.3, 51 downloads) all reconfirmed. | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------ | | Version / pins | v0.6.3 (2026-06-24); mq-lang/mq-markdown 0.6.3; reuse verdict pins `=0.6.3`; mq-mcp 0.1.20; 952★ ([25 — sem and mq](/reference/research/token-optimization-tools/25-sem-and-mq-structural-and-markdown-retrieval/)) | v0.6.4 (2026-07-03); mq-lang 0.6.4 + mq-markdown 0.6.4 on crates.io; mq-mcp 0.1.21; 955★. Pin recommendation should move to `=0.6.4`; note the new default-off `http-import-ureq` feature that must also stay off in capsules, alongside `http-import` | [https://github.com/harehare/mq/releases/tag/v0.6.4](https://github.com/harehare/mq/releases/tag/v0.6.4) ; [https://crates.io/api/v1/crates/mq-lang](https://crates.io/api/v1/crates/mq-lang) | minor | New in v0.6.4, worth noting for the dossier's token-budgeted retrieval pattern: CLI `--count`/`-c` match-count flag and `--skip`/`--limit` result pagination; a new mq-lint crate (LintRule trait + six rules); jq-parity builtins (walk, type filters, entries conversion, regex helpers, plus semver\_satisfies/uuid/rand families); mq-check pattern-matching against markdown node kinds; `@parametrize` in mq-test; a live playground embedded on mqlang.org (PR #1966). The hub's local measurements (53× TOC, 43× section extraction) were made with the v0.6.3 binary; no reason to believe 0.6.4 changes them, but they are now against a superseded release. ## Observability set — deltas [#observability-set--deltas] Core recorded facts in chapter [23 — Token observability and session visualization (token-optimizer and peers)](/reference/research/token-optimization-tools/23-token-observability-and-visualization/) all hold: PolyForm Noncommercial 1.0.0 license (the GitHub API reports SPDX NOASSERTION because PolyForm is not auto-detected; the LICENSE file confirms), JSONL-to-SQLite architecture, localhost dashboard, keep-warm pinger still opt-in and API-billed-scoped. Most load-bearing: **the thinking blind spot holds** — the README mentions no thinking-vs-visible split and GitHub code search returns 0 hits for `count_tokens` in token-optimizer and claude-usage and no thinking-estimator code in token-dashboard, so the hub's "gap worth a jackin❯ tool" claim stands (caveat: code search may under-index small repos; corroborated by README content but not an exhaustive source-tree audit). | Topic | Recorded | Latest (2026-07-05) | Source | Impact | | -------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------ | | token-optimizer version/activity | "v5" layer; pushed 2026-06-15; 1,341★ / 110 forks ([23 — Token observability and session visualization](/reference/research/token-optimization-tools/23-token-observability-and-visualization/), 06-15) | v5.11.35 (2026-07-03; five patch releases that day alone); 1,542★ / 8 watchers / 122 forks / 1 open issue; ratio \~193:1 — PR-inflation signal unchanged in kind | [https://api.github.com/repos/alexgreensh/token-optimizer](https://api.github.com/repos/alexgreensh/token-optimizer) | minor | | Claimed savings | Monthly tiers: $80–150 light / $300–600 heavy / $1,500–2,500 high-waste ([23 — Token observability and session visualization](/reference/research/token-optimization-tools/23-token-observability-and-visualization/)) | README now headlines \~$313/mo counted/metered and \~$1,877/mo (\~18%) as a 30-day counterfactual over 684 sessions, plus a "75% of context waste" coverage claim vs competitors' 15–25%; the three-tier figures are gone | [https://raw.githubusercontent.com/alexgreensh/token-optimizer/HEAD/README.md](https://raw.githubusercontent.com/alexgreensh/token-optimizer/HEAD/README.md) | minor | | token-dashboard | 584★ / 11 watchers, live 30 s-refresh dashboard ([23 — Token observability and session visualization](/reference/research/token-optimization-tools/23-token-observability-and-visualization/)) | 619★ / 11 watchers / 164 forks / 20 open issues; **last push 2026-04-20** — \~2.5 months dormant, growing issue count, no releases. Functionally as described but the "#1 pure-observability pick" ranking deserves a maintenance caveat | [https://api.github.com/repos/nateherkai/token-dashboard](https://api.github.com/repos/nateherkai/token-dashboard) | minor | | claude-usage | 1,826★ / 16 watchers, MIT ([23 — Token observability and session visualization](/reference/research/token-optimization-tools/23-token-observability-and-visualization/)) | 1,952★ / 16 watchers / 360 forks; v1.5.4 released 2026-07-01, actively maintained | [https://api.github.com/repos/phuryn/claude-usage/releases/latest](https://api.github.com/repos/phuryn/claude-usage/releases/latest) | minor | New mechanisms in v5.11.35: Total Recall coordination (defers low-relevance continuity injection to the external memory tool when detected, relevance bar raised to 0.6 — an anti-double-memory-layer guard, a pattern the hub's stacking chapter cares about); WSL-root Windows home recovery for Copilot/OpenCode installs; README-level progressive disclosure (archiving large tool results), decision extraction, activity-mode detection, and a post-compaction "context intel digest" — extending the smart-compaction checkpoint mechanism already mapped in the records ledger. ## No material change [#no-material-change] * **fff** (`dmtrKovalenko/fff`): last push 2026-07-02 — nothing moved since the 07-03 sweep. All load-bearing facts checked: v0.9.6 latest stable and crates.io `max_stable_version` for fff-search/fff-query-parser; nushell still pins `fff-search = "=0.9.6"` (live Cargo.toml check); the agent-formatting/MCP layer still bin-only (crates `fff-mcp` and `fff-core` do not exist on crates.io); author-controlled forks (neo\_frizbee 0.10.3, fff-notify-debouncer-full 0.9.4, zlob `=1.6.0-dev.7`), vendored libgit2, heed 0.22, and the notify RC pin (rc.3 pinned; upstream now at rc.4, stable still 8.2.0 — the docs' `notify = "8"` recommendation holds) all confirmed in the workspace manifest; the crates.io `fff` name-squat caveat unchanged; T1-latency/T4-token verdict in chapter [19 — Code-intelligence tools: codedb, fff, CodeGraff, and alternatives](/reference/research/token-optimization-tools/19-code-intelligence-tools/) stands. Two footnotes: a **third** published crate exists — fff-grep 0.9.6, a workspace dependency resolved transitively, so the "two crates" framing is slightly incomplete (minor, not an error) — and 0.9.7 is imminent (daily `0.9.7-nightly.*` prereleases through 2026-07-02; PR #645 rewrites exactly the per-platform file walker and gitignore matcher that the "default-features off → pure-Rust walker" guidance depends on, so the `=0.9.6` pin needs a deliberate re-check at 0.9.7). Stats: 9,524★ / 20 watchers / 367 forks / 45 open issues ([repo API](https://api.github.com/repos/dmtrKovalenko/fff), [fff-grep](https://crates.io/api/v1/crates/fff-grep)). * **tokbench** (`Entelligentsia/tokbench`): no R2 — last pushed 2026-06-15, no releases, README still "pilot, N=1 per arm — replication runs in progress"; the R1 numbers the hub quotes are confirmed verbatim in the live README ([repo API](https://api.github.com/repos/Entelligentsia/tokbench)). * **cavegemma**, **issue rtk#1155**, **RTK Cloud waitlist**, **headroom Phase H**, **sem cli/mcp publishing wedge**: all explicitly re-checked and unchanged (sources in the sections above). ## Corrections register — chapters 01–16 and 19–25 [#corrections-register--chapters-0116-and-1925] Chapters 13 and 15 stay frozen as dated records; this table is the authoritative correction layer on top of them. "Chapter" names the page carrying the now-stale claim. | Chapter | Stale claim | Correction (2026-07-05) | Source | | ------- | ----------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | 01 | Facts table "Latest seen: v1.9.0"; adoption 74,446★ / 166 w (06-18); pitch row "README headline cuts \~75%" | v1.9.1 (2026-07-03, "65%, honestly"); 83,839★ / 194 w; headline now flat "65% average output reduction (range 22–87%)" — \~75% formally retired in v1.9.1 (#234) | [https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1](https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1) | | 01 | "Evidence and claims to kill" bullet: "Caveman cuts \~75% of tokens" | Reframe: the project itself retired \~75% and now claims 65% output-only average, and ships `docs/HONEST-NUMBERS.md` conceding when caveman costs tokens; the hub's 58.5% Claude-tokenizer replication and \~17%-of-dollars ceiling remain the deeper correction | [https://github.com/JuliusBrussee/caveman/blob/v1.9.1/docs/HONEST-NUMBERS.md](https://github.com/JuliusBrussee/caveman/blob/v1.9.1/docs/HONEST-NUMBERS.md) | | 14 | TL;DR: "No release since v1.9.0 (2026-06-12), yet +6.8k stars in 15 days of zero commits" | v1.9.1 shipped 2026-07-03 (50 changes; claim rebase to 65%, hook/installer hardening, HONEST-NUMBERS.md); the promotion-driven-adoption signal now rests on star velocity alone, not on a frozen repo | [https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1](https://github.com/JuliusBrussee/caveman/releases/tag/v1.9.1) | | 14 | cavemem: "last push 2026-05-17 — the entire project lived one month"; "44 open issues, zero maintainer replies" | cavemem resumed maintenance 2026-07-03: issues 44 → 4, npm packaging bug fixed, opt-in web-enrichment MCP tool, PowerShell 7 hooks, in-repo 0.3.0 (npm still 0.2.1). The measurement-based "refuted as shipped" critique stands until re-measured, but the unmaintained claim is false | [https://github.com/JuliusBrussee/cavemem/commits/main](https://github.com/JuliusBrussee/cavemem/commits/main) | | 14 | Family growth: "cavegemma (73 stars)", "cavekit 1,068 stars", "caveman-shrink 27k npm dl/mo" | cavegemma 79★ (still frozen); cavekit 1,074★, now at v4.1.0 (tag, 2026-06-18); caveman-shrink 23.3k dl/mo, and its published tarball was missing `spawn-options.js` until fixed alongside v1.9.1 (#597) | [https://api.github.com/repos/JuliusBrussee/cavekit/tags](https://api.github.com/repos/JuliusBrussee/cavekit/tags) | | 02 | Summary table: repo `chopratejas/headroom`, "Latest seen v0.26.0 (\~190 PyPI releases)", 33,359★ / 111 w, companion model `kompress-base` | Repo `headroomlabs-ai/headroom` (redirect intact); v0.30.0 (195 PyPI releases); 56,509★ / 176 w; companion model kompress-v2-base (ModernBERT-base 149M), still on the founder's personal HF account | [https://api.github.com/repos/headroomlabs-ai/headroom](https://api.github.com/repos/headroomlabs-ai/headroom) ; [https://pypi.org/pypi/headroom-ai/json](https://pypi.org/pypi/headroom-ai/json) | | 13 | headroom bullet "(v0.28.0 …)"; ledger "PyPI/npm 0.28.0"; "New `--lossless` proxy mode in Unreleased" | v0.29.0 and v0.30.0 both shipped 2026-07-03, after the sweep's clone; `--lossless` is released (v0.29.0); note `--force-kompress-all` as an opt-in counterweight to the ML-demotion narrative | [https://github.com/headroomlabs-ai/headroom/releases](https://github.com/headroomlabs-ai/headroom/releases) | | 13 | RTK "Develop (0.44-rc)" list includes "`uv run` transparent prefix" | Removed from `BUILTIN_TRANSPARENT_PREFIXES` (commit `96f9422`, PR #2768, 2026-07-01) to fix rewrite conflicts; the incoming-0.44 list gains instead: TOML filters matched in the PreToolUse hook via RegexSet (PR #2748), trust-gating extended to `rtk trust` with project/global scoping | [https://github.com/rtk-ai/rtk/commit/96f9422](https://github.com/rtk-ai/rtk/commit/96f9422) ; [https://github.com/rtk-ai/rtk/pull/2748](https://github.com/rtk-ai/rtk/pull/2748) | | 15 | "headroom `c75ebde` (no commits since)" | v0.29.0 + v0.30.0 released 2026-07-03 plus post-release fixes (HEAD `e8151f0`, 2026-07-03T22:20Z) | [https://github.com/headroomlabs-ai/headroom/releases/tag/v0.29.0](https://github.com/headroomlabs-ai/headroom/releases/tag/v0.29.0) | | 15 | "lean-ctx `a80350da` == v3.8.18 (no commits since)" | \~30 commits since, including the v3.9.0 version bump and CHANGELOG cut (2026-07-04); v3.9.0 unreleased (no tag; crates.io still 3.8.18) — the clause should be read as scoped to the 2026-07-03 sweep date | [https://github.com/yvgude/lean-ctx/commits/main](https://github.com/yvgude/lean-ctx/commits/main) ; [https://crates.io/api/v1/crates/lean-ctx](https://crates.io/api/v1/crates/lean-ctx) | | 15 | codedb: "current is 0.2.5826, not 5827"; "history squashed to one commit; no changes since 2026-06-23" | Latest GitHub release is v0.2.5827 (2026-06-23); main HEAD remains `c12e1e2` == v0.2.5826 so all main-anchored install/telemetry evidence still holds, but 0.2.5828 is in active preparation (merges through 2026-07-04) and main shows 100+ commits of history | [https://github.com/justrach/codedb/releases/tag/v0.2.5827](https://github.com/justrach/codedb/releases/tag/v0.2.5827) | | 03 | Header table: "Latest seen v0.42.4 (212 release tags)"; adoption 63,608★ / 146 w (06-18) | v0.43.0 (2026-06-28); develop at 0.44-rc (`dev-0.44.0-rc.308`); 68,484★ / 166 w / 4,232 forks / 1,485 open issues — still PR-inflated (\~412:1) | [https://api.github.com/repos/rtk-ai/rtk/releases/latest](https://api.github.com/repos/rtk-ai/rtk/releases/latest) | | 04 | Fact table: "Latest seen v3.8.9"; adoption (06-20) 2,800★ / 19 w / 278 forks / 13 issues | Latest released v3.8.18 (2026-07-01; v3.9.0 cut in-repo, unreleased); 3,085★ / 21 w / 294 forks / 34 open issues. Once v3.9.0 ships, the "what it lacks"/commercial-model sections need the AI Gateway team mode and the grammar-addon dylib mechanism | [https://api.github.com/repos/yvgude/lean-ctx](https://api.github.com/repos/yvgude/lean-ctx) ; [https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/yvgude/lean-ctx/HEAD/CHANGELOG.md) | | 21 | Table: repo `chopratejas/headroom`, v0.25.0, 28,199★, kompress-base | Page carries explicit "accessed 2026-06-15" dating, so lower priority; if refreshed: `headroomlabs-ai/headroom`, v0.30.0, 56,509★ / 176 w / 4,128 forks, model kompress-v2-base. Otherwise keep as the dated 06-15 snapshot — the hub's canonical current-state pages are 02/13/15 plus this chapter | [https://api.github.com/repos/headroomlabs-ai/headroom](https://api.github.com/repos/headroomlabs-ai/headroom) | | 25 | sem: "sem-core 0.16.2 is a real crates.io library"; "\~50 releases to v0.16.2 in 5 months"; pin `sem-core = "=0.16.2"` | v0.18.0 / sem-core 0.18.0 (2026-07-03); pin should read `=0.18.0` or note two breaking 0.x minors since 0.16.2 (0.16.0 added public `ContextResult.omitted`, breaking struct-literal construction; 0.17/0.18 add unique-method-name edges, `Parent::child` qualifiers, `orient --pack`) | [https://crates.io/api/v1/crates/sem-core](https://crates.io/api/v1/crates/sem-core) | | 25 | sem: "2.3× more accurate agents … circular n=3 self-benchmark … cite the mechanism, not the number (T4)" | Recast in past tense: sem retired the 2.3×/75% hero claims; benchmarks page now reports paired-run SWE-bench Verified A/B (50–65% faster code understanding, 2.90 s → 0.59 s/iter verify loop when edges resolve, bimodal 1.2× floor disclosed, token parity and unchanged solve rate stated plainly) | [https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md) | | 25 | sem interception list: "`sem setup` rewiring `git diff` itself … not adopted" | Incomplete: `sem setup` now also edits `~/.claude/settings.json` (SessionStart resident server + UserPromptSubmit injection hooks, unreleased on main); sem is on the official MCP registry as `io.github.Ataraxy-Labs/sem`. Strip verdict unchanged | [https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md](https://raw.githubusercontent.com/Ataraxy-Labs/sem/HEAD/CHANGELOG.md) | | 25 | mq: "mq-lang 0.6.3 + mq-markdown 0.6.3", "v0.6.3 (2026-06-24), 952 stars"; pins `=0.6.3` | v0.6.4 (2026-07-03); both crates 0.6.4; 955★; mq-mcp 0.1.21; mq-markdown \~84k downloads; pins `=0.6.4`; http-import guidance now covers the new default-off `http-import-ureq` feature too | [https://github.com/harehare/mq/releases/tag/v0.6.4](https://github.com/harehare/mq/releases/tag/v0.6.4) | | 23 | token-optimizer table: adoption/activity rows (06-15); savings tiers $80–150 / $300–600 / $1,500–2,500 | 1,542★ / 8 w / 122 forks / 1 issue; pushed 2026-07-03; v5.11.35. README now claims \~$313/mo counted and \~$1,877/mo (\~18%) counterfactual (30 days / 684 sessions), plus 75% context-waste coverage vs competitors' 15–25% | [https://raw.githubusercontent.com/alexgreensh/token-optimizer/HEAD/README.md](https://raw.githubusercontent.com/alexgreensh/token-optimizer/HEAD/README.md) | | 23 | Landscape rows for token-dashboard (584★) and claude-usage (1,826★); "#1 pure-observability pick" ranking | token-dashboard 619★ / 11 w, last push 2026-04-20 — dormant \~2.5 months, 20 open issues, no releases (maintenance caveat on the #1 pick); claude-usage 1,952★ / 16 w, v1.5.4 (2026-07-01), active | [https://api.github.com/repos/nateherkai/token-dashboard](https://api.github.com/repos/nateherkai/token-dashboard) | | 19 | codedb "npx launcher" interface row | npm `codedeebee` latest is 0.2.5823 (2026-05-29), four releases behind GitHub v0.2.5827; the release-notes-advertised `codedeebee@0.2.5827` does not exist on npm | [https://registry.npmjs.org/codedeebee](https://registry.npmjs.org/codedeebee) | ## New-tools sweep (2026-06/07 window) [#new-tools-sweep-2026-0607-window] Method: GitHub search over token/context/claude/prompt-cache keyword combinations restricted to repos created after 2026-05, plus web search over HN, Show HN, and June–July 2026 roundups; watcher/fork ratios verified via the repos API since search results only echo stargazers. Market context worth one sentence in the hub's market chapter: Anthropic's 2026-06-15 subscription split (separate API-credit pool for Agent SDK / headless usage) is visibly driving a fresh wave of these tools. Net result: **only one genuine major arrival (Recall)**; the rest are watch-list or skeptic-exhibit grade. | Name | Repo | What it is | Stats (2026-07-05) | Disposition | | ----------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Recall | raiyanyahya/recall | Claude Code plugin: durable per-project memory via hooks that log sessions and a local TF-IDF+TextRank summarizer producing a \~1–2K-token `context.md` resume — zero model tokens spent on the memory itself | 674★ / 40 forks / 2 w; Python; created 2026-06-19, pushed 2026-06-26 | The one genuine breakout: kills cold-start re-explanation with a classical non-LLM local summarizer — a mechanism distinct from cavemem/headroom. Thin watcher ratio but 40 forks + press suggest a real HN spike. Dossier candidate | | claude-pulse | nikitadoudikov/claude-pulse | Local zero-dependency dashboard: live token/context usage, lost-session recovery, transcript search, phone-based tool-call approval | 116★ / 4 forks / 0 w; JS; created 2026-06-20 | New observability-class entrant; 0 watchers on 116 stars is the inflated-ratio signature — treat the star count as unverified. Watch list | | tokdiet | agiwhitelist/tokdiet | Streaming reverse proxy (Claude Code/Cursor/Codex) metering every token/USD, compacting context, and running a **shadow-eval** to prove output quality held after compaction | 37★ / 2 forks / 0 w; TS; created 2026-06-16, pushed 2026-06-18 | The shadow-eval is the mechanism the hub's verification page names as missing from most tools — but no commits since 06-18; possibly abandoned two days after launch. Watch list | | context-forge | eatakishiyev/context-forge | Self-described "context compiler": scores, compresses, reorders, and budgets everything entering the window (drop-in proxy + reproducible benchmark), positioned against context rot | 7★ / 0 forks / 1 w; Python; created 2026-06-27, pushed 2026-07-04 | Tiny but honest and actively developed; the compiler/linter framing (budgeting + reordering, not just deletion) is architecturally the closest thing in the sweep to the [jackin❯ Context Engine spec](/reference/research/jackin-context-engine/). Watch-list row, not a dossier | | Agent Blackbox | TaewoooPark/Agent-Blackbox | Local-first flight recorder: replays runs as a session map, scores the "context bill" per run, writes the derived fix back into AGENTS.md | 44★ / 6 forks / 1 w; TS; created 2026-06-16, pushed 2026-06-29 | Novel feedback loop — attributes token cost to agent behaviors and patches the instructions file so the spend does not recur; no direct analogue in the hub taxonomy. Watch list | | claude-mega-brain | guhcostan/claude-mega-brain | Injects a project knowledge base into every Claude Code session, marketed adjacent to token savings | 109★ / 13 forks / 0 w; Python; created 2026-06-29 | Skeptic exhibit only: 109 stars in \~48 h with 0 watchers is the classic PR-inflation pattern, and per-session KB injection *increases* baseline input tokens. Candidate for the "looks like optimization, isn't" bucket | Rejected below note-level: iKislay/copium (8★, "65–90% savings zero quality loss" hype, no traction), tokenline (16★, cache-TTL statusline), the Graphify ecosystem (underlying repo predates the window; the 820★ memory-setup repo is April-era), MohammedBasioni/scalpel (cache-aware context-surgery proxy, 2★ — interesting idea, zero traction). Third-party roundups (Pinggy "8 tools", firecrawl "12 ways") surfaced nothing outside existing hub coverage except Graphify and prompt-caching basics — an explicit negative result for the "did we miss a major tool" question. ## pxpipe joins the comparison set [#pxpipe-joins-the-comparison-set] pxpipe (`teamchong/pxpipe`, MIT, TypeScript; v0.8.0 on npm as `pxpipe-proxy`, published 2026-07-04; 1,728★ / 9 watchers / 100 forks, created 2026-05-20) enters the hub as a **new category**: cache-aligned prompt re-encoding ("optical context compression") — a local proxy that rewrites the static system/tool-docs slab, collapsed older history, and large tool\_results into dense PNG pages to exploit flat vision-token pricing, while relocating rather than adding cache markers so the prefix cache stays warm. It is the first covered tool that directly attacks the cache-read bucket, and its per-request `count_tokens`-counterfactual accounting is the anti-K1 gold standard; it is also lossy with a silent-confabulation failure shape and its economics are hostage to provider pricing/vision behavior. Full dossier, claims table, risk register, and engine recommendation (PORT the measurement kernel and cache-alignment invariants; do not adopt the imaging mechanism yet): [17 — pxpipe design](/reference/research/token-optimization-tools/17-pxpipe-design/). Naming footnote for future sweeps: the npm package is `pxpipe-proxy`; the bare npm name `pxpipe` is unclaimed (typo-install squat risk), and GitHub Releases lag the git tags (only v0.7.0/v0.7.1 have release objects) — npm is the authoritative artifact stream. ## Verification limits [#verification-limits] Negative results and fetch failures this sweep, recorded so the next sweep does not re-derive them: * cavemem 0.3.0 exists only in the repo `package.json`; the npm registry still serves 0.2.1 as latest — the publish is pending, not verifiable. caveman-code and cavegemma have no GitHub releases or tags (404 on releases/latest) — expected; their versioning is npm-only/none. * huggingface.co API for `headroomlabs-ai/kompress-v2-base` returned an auth error — interpreted as the model being absent under the org (the personal-account location was confirmed positively), but a private org mirror cannot be ruled out from the API alone. * crates.io rejects the default curl User-Agent per its data-access policy; all crates.io reads in this sweep succeeded on retry with an identifying UA — noted so future sweeps do not misread the first-attempt failure as an outage. * RTK's 77-command/62-module/86-rule counts were not independently recounted (they need a clone-level audit like the 07-03 sweep); only the 63-TOML-filter count was checked via the git tree API. No contrary evidence found. * lean-ctx issue #625 ("recovery-first turn"): GitHub `yvgude/lean-ctx#625` is an unrelated closed hook bug; the changelog cites two trackers (GH and GL numbering) and the recorded #625 is likely GitLab-side with no discoverable URL — the recovery-first-turn citation could be confirmed neither way. The faithful-arm "minimal profile" note was likewise not checked again (not contradicted, just unchecked). No v3.9.0 artifact exists, so the 75.9 MiB binary figure could not be re-measured. * codedb runtime claims (5.16k-token MCP schema, 1.7 GB RSS, telemetry upload) were confirmed only by proxy via the unchanged main HEAD, not re-measured live. token-optimizer ships no CHANGELOG file (404), so v5.x deltas were reconstructed from GitHub release notes only. * mq's \~291-crate dependency-tree count was not re-resolved (needs `cargo metadata`); direction verified via Cargo.toml showing format deps still non-optional. sem's `docs/attention-architecture.md` was not re-fetched; ledger status was inferred from changelog absence of ledger changes. ## Ledger [#ledger] All accessed 2026-07-05. | Source | Used for | | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `api.github.com/repos/JuliusBrussee/{caveman, cavemem, caveman-code, cavekit, cavegemma}`; users/JuliusBrussee/repos?sort=created; github.com/JuliusBrussee/caveman/releases/tag/v1.9.1; …/caveman/blob/v1.9.1/docs/HONEST-NUMBERS.md; …/caveman/issues/484; …/cavemem/commits/main; …/cavekit/tags; api.npmjs.org/downloads/point/last-month/caveman-shrink | caveman-family versions, stats, release contents, claim rebase, cavemem revival, npm download counts | | api.github.com/repos/headroomlabs-ai/headroom (+ /languages, /commits, /contents/headroom/proxy/server.py); github.com/headroomlabs-ai/headroom/releases/tag/v0.29.0 and .../v0.30.0; raw README.md + REALIGNMENT/10-phase-H-python-retirement.md; pypi.org/pypi/headroom-ai/json; `crates.io/api/v1/crates/{headroom-core, headroom}`; huggingface.co/api/models/chopratejas/kompress-v2-base | headroom versions, release contents, Phase H status, crate squat, model location, headline-claim check | | api.github.com/repos/rtk-ai/rtk (+ /releases/latest, /git/trees/develop?recursive=1); `github.com/rtk-ai/rtk/{commit/96f9422, pull/2748, issues/1155, releases}`; raw develop Cargo.toml, CHANGELOG.md, README.md; [www.rtk-ai.app](http://www.rtk-ai.app) | RTK version, filter recount, uv-run revert, hook RegexSet matching, Cloud waitlist, docs drift, #1155 | | api.github.com/repos/yvgude/lean-ctx (+ /releases/latest, /commits/main); raw CHANGELOG.md; crates.io/api/v1/crates/lean-ctx; api.github.com/repos/Entelligentsia/tokbench + raw README.md | lean-ctx release state, v3.9.0 contents, corruption incidents, tokbench R1 confirmation / no-R2 | | api.github.com/repos/justrach/codedb (+ /commits?sha=main, /compare/main...release/0.2.5828, /activity); `github.com/justrach/codedb/{releases/tag/v0.2.5827, commits/release/0.2.5828}`; registry.npmjs.org/codedeebee | codedb version inversion, provenance, Windows port, opt-out revert, npm launcher lag | | api.github.com/repos/Ataraxy-Labs/sem; github.com/Ataraxy-Labs/sem/releases/tag/v0.18.0; raw CHANGELOG.md + README.md; `crates.io/api/v1/crates/{sem-core, sem-core/0.18.0, sem-cli, sem-mcp, sem-cloud-client}` | sem versions, hero-claim retirement, setup host-write growth, MCP registry listing, publishing wedge | | api.github.com/repos/dmtrKovalenko/fff; github.com/dmtrKovalenko/fff/releases/tag/v0.9.6 + main Cargo.toml + crates/fff-mcp/Cargo.toml; github.com/nushell/nushell/blob/main/Cargo.toml; `crates.io/api/v1/crates/{fff-search, fff-query-parser, fff-grep, fff-mcp, fff-core, notify}`; raw README.md | fff pin currency, nushell pin, bin-only MCP layer, fork/RC pins, nightlies, third crate | | api.github.com/repos/harehare/mq; github.com/harehare/mq/releases/tag/v0.6.4; raw crates/mq-lang/Cargo.toml; nushell nu-command Cargo.toml; `crates.io/api/v1/crates/{mq-lang, mq-markdown, mq-db, mq-mcp}` | mq version bump, format-gating check, nushell dep, prototype statuses | | api.github.com/repos/alexgreensh/token-optimizer (+ /releases/latest); raw README.md; github.com/alexgreensh/token-optimizer/blob/main/LICENSE.md; api.github.com/repos/nateherkai/token-dashboard; api.github.com/repos/phuryn/claude-usage (+ /releases/latest); GitHub code search (count\_tokens / thinking estimators) | observability-set versions, license, claim restructure, dormancy, thinking-blind-spot negative result | | api.github.com/repos/teamchong/pxpipe (+ /releases, /tags, /contributors, /commits); raw README.md, CHANGELOG.md, package.json, FINDINGS.md, docs/CACHING\_AND\_SAVINGS.md, docs/HISTORY\_CACHE\_MODEL.md, `src/core/{proxy, transform, render, history, applicability, atlas}.ts`; `registry.npmjs.org/{pxpipe-proxy, pxpipe}`; trendshift.io/repositories/71973; x.com/kimmonismus/status/2073315097317871850; aiweekly.co pxpipe alert | pxpipe identity, version, mechanism source-read, virality assessment (details in [17 — pxpipe: design teardown](/reference/research/token-optimization-tools/17-pxpipe-design/)) | | GitHub search API (token/context/claude/prompt-cache, created:>2026-05); news.ycombinator.com/item?id=47581701; pinggy.io blog "tools to reduce AI coding agent token usage"; gist.github.com/yurukusa/7d854616809e673ca8d23353ed8267a6; medium.com/@abdulgafoorabid "how I cut Claude Code token usage"; github.com/raiyanyahya/recall; repos API for each sweep candidate | new-tools sweep candidates, ratio verification, rejected list, negative result |