🇰🇷 한국어 버전 | 🔧 Quick fix guide → — skip the analysis, just fix it
Claude Code Hidden Problem Analysis
TL;DR: Claude Code has 11 confirmed client-side bugs (B1-B5, B8, B8a, B9, B10, B11, B2a) plus 3 preliminary findings (P1-P3). Cache bugs (B1-B2) are fixed in v2.1.91. Nine remain unfixed as of v2.1.119 (latest). Proxy data now covers 45,884 requests over 22 days (April 1–22, 320 unique sessions). A controlled GrowthBook flag override eliminated B4/B5 events completely (167,818 → 0, 5,500 → 0). The 7d quota window can become the binding constraint — first observed when 7d utilization hit 0.97. Anthropic acknowledged B11 (adaptive thinking zero-reasoning) on HN but has not followed up.
⚠️ Opus 4.7 advisory: Do not upgrade past v2.1.109 — 2.4x Q5h burn, model pin bypass, cache metering anomaly. Anthropic published an April 23 postmortem admitting 3 product-layer bugs; CHANGELOG analysis shows 2 of 3 were never documented — see 17_OPUS-47-POSTMORTEM-ANALYSIS.md.Last updated: April 24, 2026 — see 17_OPUS-47-POSTMORTEM-ANALYSIS.md (new: postmortem cross-check, CHANGELOG transparency analysis, post-postmortem issues), CROSS-VALIDATION-20260422.md, 16_OPUS-47-ADVISORY.md, and 08_UPDATE-LOG.md.
New chapter: 17_OPUS-47-POSTMORTEM-ANALYSIS.md — Anthropic's April 23 postmortem admitted three product-layer bugs (effort default downgrade, thinking cache clearing bug, verbosity system prompt). Cross-checking against the CHANGELOG (3,285 lines, all versions v2.1.68–v2.1.119 searched) reveals: 2 of 3 bugs have zero CHANGELOG documentation — the thinking cache bug (v2.1.101 fix) and verbosity prompt (v2.1.116 revert) were introduced and removed without any public record. The effort default change was documented but framed as a product improvement, never as a regression.
Post-postmortem issues: 5 new issues on v2.1.117–119 demonstrate problems beyond the postmortem's scope: subagent model pin ignored (#52502 — Haiku pin silently runs Opus, $10.87 vs $0.0005), effort override bypass (#52534 — unpinOpus47LaunchEffort flag), auto-compact 5x (#52522), self-conversation safety issue (#52228), CLAUDE.md rule violation (#52652). v2.1.109 recommendation remains valid. 36 claims cross-checked (28 confirmed, 5 partially confirmed, 3 not relied upon).
New document: CROSS-VALIDATION-20260422.md — Three independent datasets (ArkNill 45.8K proxy, @seanGSISG 215K JSONL, @cnighswonger 101K interceptor) converge on CacheRead per 1% utilization: 1.5–2.1M tokens across different accounts, plan tiers (Max 5x and 20x), geographies, and collection methods. Quota multiplier (0x vs 1x) ranges from 9.8x to 38.6x depending on cache-read ratio. ArkNill April multiplier: 32.9x. cnighswonger's January baseline (474 calls, >20x) confirms the weight change affects all usage levels. Proxy dataset expanded to 45,884 requests across 320 sessions; model substitution check at 41,306 requests — zero mismatches. New Q7d analysis: 13.5% of requests in 80–100% bucket vs Q5h's 0.6% — the 7-day window is the binding constraint for sustained heavy users. Before-data limitation in 02_RATELIMIT-HEADERS.md marked as resolved. Issue #3 cross-validation →
Recommendation: Stay on v2.1.109 — sends explicit claude-opus-4-6 model IDs (self-verified), has native 1h cache, and is unaffected by the April 23 API default switchover. Pin your version with DISABLE_AUTOUPDATER=1.
Self-measured benchmark: 3 effort levels × 3 hard tasks × 20-turn session isolation (n=3). Evidence quality assessment with gap analysis included in the advisory.
Proxy dataset expanded to 38,996 requests across 272 sessions (April 1–16). Full data audit: 3 labeled datasets (4,593 JSONL files / 512,149 messages / ~1.9 GB) now indexed in an internal database. New chapters: 14_DATA-SOURCES.md (label matrix + historical reconciliation) and 15_ENV-BREAKDOWN.md (per-environment cache_read, model dispatch, tier-dependent Haiku findings). Post-April 10 cache_read: ubuntu-1-override 97.08% vs ubuntu-1-stock 96.00% — consistent with 1h TTL preservation under the override. Max 5x Haiku share 0.11% vs Max 20x ~21% (190× difference).
Proxy dataset expanded to 35,554 requests across 251 sessions (April 1–15). Verified through CC v2.1.108.
1. Independent corroboration. @seanGSISG contributed a 179K-call dataset (Dec 2025 – Apr 2026, Max 20x) with 4 analysis scripts. Key results: CacheRead per 1% at 1.62-1.72M (within our 1.5-2.1M range), thinking token contribution estimated at 0.0-0.1% from JSONL content blocks, and a counterfactual showing zero days exceed budget under 0x formula while 18 days exceed under 1x. This resolves our "No before-data" limitation. Issue #3 →
2. Thinking token status update. "Blind spot" revised to "partially measured" based on @seanGSISG's JSONL analysis. Server-side computation cost remains unmeasurable from client side, but content block text suggests <1% of quota. Details →
3. Cache efficiency. Overall cache efficiency improved to 98.3% (from 97.0% at 30K requests). Post-barrier (flag override active): 9,996 requests with sustained zero B4/B5 events.
Key updates from April 14:
1. GrowthBook flag override — controlled elimination test. Deployed a proxy-based flag override on April 10 (the approach documented in #42542). Result: B5 events 167,818 → 0, B4 events 5,500 → 0 across 4,919 subsequent requests over 4 days. Same machine, same account, same usage patterns. This is the strongest causal evidence that these flags directly control context mutation. Methodology →
2. seven_day bottleneck — first observation. Previously reported representative-claim = five_hour in 100% of requests. With the expanded dataset, 22.6% of requests (5,279/23,374) showed seven_day as the binding constraint — concentrated on April 9–10 when 7d utilization reached 0.85–0.97. After the weekly reset, five_hour resumed. The 7d window is not cosmetic. Details →
3. Data interpretation caveat. The measurement environment changed on April 10 when we deployed the flag override. All B4/B5 event counts (167,818 and 5,500) are from the unmodified baseline period (April 1–10, 25,558 requests). Data from April 11 onward (4,919 requests) reflects the overridden environment. Rate limit header analysis and fallback-percentage data are unaffected by the override. Caveat →
4. Updated metrics: fallback-percentage expanded to 23,374 requests — still 0.5 on every single one, zero variance. First-turn cache miss: 77.8% (158 sessions, slightly improved from 79.0%/143 sessions).
Caught up on v2.1.98 and v2.1.101 (v2.1.99/100 don't exist — skipped in the public changelog). Two more releases, still zero fixes for B3–B11. v2.1.98 was mostly security patches (Bash permission bypasses). v2.1.101 fixed resume and MCP bugs — B2a (SendMessage cache miss) may be fixed via the CLI resume path, but the Agent SDK code path is unconfirmed. Changelog cross-reference →
The "Output efficiency" system prompt section (P3) appears to be gone. Scanned all 353 local JSONL session files — every session after April 10 shows zero occurrences of the "straight to the point" / "do not overdo" text. First noticed by @wjordan via system prompt archive diffing. P3 update →
Also measured first-turn cache performance across 143 sessions (≥3 requests each): 79% start with cache_read=0 on the first API call, even on v2.1.91+ where B1/B2 are fixed. This is structural — skills and CLAUDE.md land in messages[0] instead of the system[] prefix, breaking prefix-based caching for new sessions. Newer versions are improving this (community data shows ~29% on v2.1.104), but it's still a significant first-turn cost. Details →
5 new bugs + 3 preliminary findings from community-wide issue/comment analysis and fact-checking (April 6-9):
| Bug | What | Evidence | Details |
|---|---|---|---|
| B8a | JSONL non-atomic write → session corruption | ~10+ duplicates in #21321 | 01_BUGS.md |
| B9 | /branch context inflation (6%→73%) |
3 duplicate issues | 01_BUGS.md |
| B10 | TaskOutput deprecation → 21x context injection → fatal | has repro |
01_BUGS.md |
| B11 | Adaptive thinking zero-reasoning → fabrication | Anthropic acknowledged (HN) | 01_BUGS.md |
| B2a | SendMessage resume: cache_read=0 (even system prompt) | cnighswonger confirmed | 01_BUGS.md |
Preliminary findings (MODERATE): P1/P2 cache TTL dual tiers — two triggers for 1h→5m downgrade: telemetry disabled (has repro) and quota exceeded. P3 "Output efficiency" system prompt (v2.1.64). P4 (third-party detection gap) removed April 14 — insufficient evidence. See 01_BUGS.md — Preliminary Findings.
Changelog cross-reference (v2.1.92–v2.1.97): Six releases shipped zero fixes for the nine unfixed bugs. See 01_BUGS.md — Changelog Cross-Reference.
April 8 — Full-week proxy dataset — 13_PROXY-DATA.md
cc-relay proxy database now covers 17,610 requests across 129 sessions (April 1-8), with automated bug detection across 532 JSONL files (158.3 MB):
| Metric | Previous (Apr 3) | Current (Apr 1-8) | Change |
|---|---|---|---|
| Budget enforcement (B5) | 261 events | 72,839 events | 279x |
| Microcompact (B4) | 327 events | 3,782 events (15,998 items) | 12x |
| B8 inflation (bulk scan) | 2.87x (1 session) | 2.37x avg (10 sessions, max 4.42x) | Universal |
| Synthetic rate limit (B3) | 24 entries / 6 days | 183/532 files (34.4%) with <synthetic> model entries |
Pervasive |
| Context growth rate | +575 tok/turn | median 1,845 tok/min (53 sessions) | Statistical |
New findings:
- Request rate: Mean 2.72 req/min across 78 sessions. Sustained max 8.04 req/min (60+ min sessions). Two very short sessions (2-3 min) averaged 12+ req/min; burst peak 86 req/60s from subagent fan-out.
- Per-request cost scales with session length: 0-30min: $0.20/req → 5hr+: $0.33/req (structural, not version-specific)
- Cache efficiency stable: 98-99% across all session lengths on v2.1.91 (Bugs 1-2 fully fixed)
- Subagent gap: Haiku 58.1% cache vs Opus 98.8% — 40pp gap persists
- Microcompact intensifies: 1.6 items/event at <10 messages → 6.6 items/event at 200+ messages
Rate limit header analysis — 02_RATELIMIT-HEADERS.md
Transparent proxy (cc-relay) captured anthropic-ratelimit-unified-* headers across 27,708 requests (April 1-13), revealing the server-side quota architecture:
Dual sliding window system:
- Two independent counters: 5-hour (
5h-utilization) and 7-day (7d-utilization) representative-claim=five_hourin 100% of requests — the 5h window is always the bottleneck- 5h windows reset on roughly 5-hour intervals; 7d resets weekly (April 10, 12:00 KST for this account)
Per-1% utilization cost (measured across 5 active windows on Max 20x / $200/mo):
| Metric | Range | Note |
|---|---|---|
| Output per 1% | 9K-16K | Visible output only (thinking excluded) |
| Cache Read per 1% | 1.5M-2.1M | 96-99% of visible token volume |
| Total Visible per 1% | 1.5M-2.1M | Output + Cache Read + Input |
| 7d accumulation ratio | 0.12-0.17 | 7d_delta relative to 5h_peak |
Thinking token blind spot: Extended thinking tokens are not included in the output_tokens field from the API. At 9K-16K visible output per 1%, a full 5h window (100%) = only 0.9M-1.6M visible output tokens — low for several hours of Opus work. The gap is consistent with thinking tokens being counted against the quota, but the exact mechanism can't be confirmed from the client side. Thinking-disabled isolation test planned for the week of April 6.
Community cross-validation:
- @fgrosswig: 64x budget reduction — dual-machine 18-day JSONL forensics (Mar 26: 3.2B tokens no limit → Apr 5: 88M at 90%)
- @Commandershadow9: 34-143x capacity reduction — cache fix confirmed, capacity drop independent of cache bug, thinking token hypothesis
v2.1.89 separation: The cache regression (Mar 28 - Apr 1) is a separate, resolved issue. The capacity reduction exists independently — clean comparison: golden period (Mar 23-27, cache 98-99%) vs post-fix (Apr 2+, cache 84-97%), both with healthy cache. Data collection ongoing through April 10 (full 7d cycle).
pie title Bug Status (12 identified, verified through v2.1.119)
"Fixed (B1, B2)" : 2
"Unfixed (B3-B5, B8-B11, B8a)" : 8
"Possibly Fixed (B2a)" : 1
"By Design (Server)" : 1
Cache regression (v2.1.89) is fixed in v2.1.90-91. Eight client-side bugs remain unfixed through v2.1.112 (latest). B2a (SendMessage resume) possibly fixed in v2.1.101 (CLI resume path fixed, SDK path unconfirmed). P3 ("Output efficiency" prompt) observed removed (self-verified). Changelog cross-reference: 01_BUGS.md § Changelog Cross-Reference.
| Bug | What It Does | Impact | Status | Details |
|---|---|---|---|---|
| B1 Sentinel | Standalone binary corrupts cache prefix | 4-17% cache read (v2.1.89) | Fixed | 01_BUGS.md |
| B2 Resume | --resume replays full context uncached |
Full cache miss per resume | Fixed | 01_BUGS.md |
| B2a SendMessage | Agent SDK SendMessage resume: full cache miss including system prompt | cache_read=0 on first resume | Possibly Fixed | 01_BUGS.md |
| B3 False RL | Client blocks API calls with fake error | Instant "Rate limit reached" | Unfixed | 01_BUGS.md |
| B4 Microcompact | Tool results silently cleared mid-session | 5,500 events, 18,858 items cleared | Unfixed | 01_BUGS.md |
| B5 Budget cap | 200K aggregate limit on tool results | 167,818 events, 100% truncation | Unfixed | 01_BUGS.md |
| B8 Log inflation | Extended thinking duplicates JSONL entries | 2.37x avg (max 4.42x), universal | Unfixed | 01_BUGS.md |
| B8a JSONL corruption | Concurrent tool execution drops tool_result → permanent 400 | ~10+ duplicates in #21321 | Unfixed | 01_BUGS.md |
| B9 /branch inflation | Message duplication/un-compaction on branch | 6%→73% context in one message | Unfixed | 01_BUGS.md |
| B10 TaskOutput thrash | Deprecation message triggers 21x context injection → fatal | 87K vs 4K, triple autocompact | Unfixed | 01_BUGS.md |
| B11 Zero reasoning | Adaptive thinking emits zero reasoning → fabrication | Anthropic acknowledged | Investigating | 01_BUGS.md |
| Server | Quota architecture + thinking token accounting | Reduced effective capacity | By design | 02_RATELIMIT-HEADERS.md |
- Update to v2.1.91+ — fixes the cache regression (worst drain). v2.1.92–108 add no bug fixes for issues tracked here but are safe to use
- npm or standalone — both fine on v2.1.91 (Sentinel gap closed)
- Don't use
--resumeor--continue— replays full context as billable input - Start fresh sessions periodically — the 200K tool result cap (B5) silently truncates older results
- Avoid
/dreamand/insights— background API calls that drain silently
See 09_QUICKSTART.md for setup guide and self-diagnosis. Full proxy dataset: 13_PROXY-DATA.md.
Even with cache at 95-99%, drain persists. At least four server-side issues contribute:
1. Server-side accounting change: Old Docker versions (v2.1.74, v2.1.86 — never updated) started draining fast recently, proving the issue isn't purely client-side (#37394).
2. 1M context billing regression: A late-March regression causes the server to incorrectly classify Max plan 1M context requests as "extra usage." Debug logs show a 429 error at only ~23K tokens (#42616).
3. Dual-window quota + thinking token blind spot: 5h + 7d independent windows. Visible output only 9K-16K per 1% — the gap is likely thinking tokens counted against quota but invisible to clients. Full analysis: 02_RATELIMIT-HEADERS.md.
4. Org-level quota sharing: Accounts under the same organization share rate limit pools. passesEligibilityCache and overageCreditGrantCache are keyed by organizationUuid, not accountUuid. Originally discovered by @dancinlife through client-side analysis of the obfuscated JavaScript bundle.
See 09_QUICKSTART.md for the full list of behaviors to avoid and adopt, including /branch, /release-notes, and environment variable recommendations.
On April 1, 2026, my Max 20 plan ($200/mo) hit 100% usage in ~70 minutes during normal coding. JSONL analysis showed the session averaging 36.1% cache read (min 21.1%) where it should have been 90%+. Every token was being billed at full price.
Downgrading from v2.1.89 to v2.1.68 immediately recovered cache to 97.6% — confirming the regression was version-specific. I set up a transparent monitoring proxy (cc-relay) to capture per-request data going forward.
What started as personal debugging quickly expanded. Dozens of users were reporting the same symptoms across what became 91+ GitHub issues. Community members — @Sn3th, @rwp65, @fgrosswig, @seanGSISG, @Commandershadow9, and others — independently found different pieces of the puzzle.
The investigation timeline:
| Date | What happened |
|---|---|
| Apr 1 | 70-minute 100% drain → v2.1.89 regression confirmed, proxy setup |
| Apr 2 | Bugs 3-4 discovered (false rate limiter, silent microcompact). Anthropic's Lydia Hallie posts on X |
| Apr 3 | Bug 5 discovered (200K budget cap). v2.1.91 benchmark: cache fixed, 4 other active bugs persist (B3-B5, B8). 06_TEST-RESULTS-0403.md |
| Apr 4-6 | cc-relay captures 3,702 requests with rate limit headers. Community analysis continues |
| Apr 6 | Dual-window quota analysis published. Community cross-validation (fgrosswig 64x, Commandershadow9 34-143x). 02_RATELIMIT-HEADERS.md |
Full 14-month chronicle (Feb 2025 – Apr 2026): 07_TIMELINE.md
Lydia Hallie (Anthropic, Product) posted on X:
"Peak-hour limits are tighter and 1M-context sessions got bigger, that's most of what you're feeling. We fixed a few bugs along the way, but none were over-charging you."
She recommended using Sonnet as default, lowering effort level, starting fresh instead of resuming, and capping context with CLAUDE_CODE_AUTO_COMPACT_WINDOW=200000.
Where our data diverges from this assessment:
- "None were over-charging you" — Bug 5 silently truncates tool results to 1-49 chars after a 200K aggregate threshold. Users paying for 1M context effectively have a 200K tool result budget for built-in tools. 261 truncation events measured in a single session.
- "We fixed a few bugs" — Cache bugs (B1-B2) are fixed, but Bugs 3-5 and B8 remain active in v2.1.91. Client-side false rate limiter (B3) generated 151 synthetic "Rate limit reached" errors across 65 sessions on our setup — zero API calls made.
- "Peak-hour limits are tighter" — Our April 6 proxy data shows the bottleneck is always the 5h window (
representative-claim=five_hourin 100% of 3,702 requests), regardless of time of day. Weekend and off-peak data shows the same pattern. - Thinking token accounting — Extended thinking tokens don't appear in
output_tokensfrom the API, yet visible output alone explains less than half the observed utilization cost. If thinking tokens are counted against quota at output-token rate, this is a significant invisible cost that users have no way to monitor or control.
GitHub response: bcherny posted 6 comments on #42796 (April 6 only, triggered by HN virality), then went silent. Zero responses on all other 90+ issues including #38335 (478 comments, 15 days). See 10_ISSUES.md for full history.
@luongnv89 documented that idle gaps of 13+ hours cause a full cache rebuild. Anthropic documents a 5-minute TTL, though our data shows 5-26 minute gaps sometimes maintaining 96%+ cache — the actual TTL may be longer in practice. Not a bug, but worth knowing about.
| File | What | Updated |
|---|---|---|
| 17_OPUS-47-POSTMORTEM-ANALYSIS.md | Postmortem cross-check: CHANGELOG transparency analysis, post-postmortem issues (v2.1.116+), effort 48-day Pro/Max gap, 36-claim verification matrix | Apr 24 |
| 01_BUGS.md | All 11 bugs (B1-B11, B2a, B8a) + 3 preliminary (P1-P3, P4 removed) + changelog cross-reference (v2.1.92-108) | Apr 15 |
| 09_QUICKSTART.md | Quick fix guide — Option A (v2.1.91+) vs Option B (v2.1.63 downgrade), npm vs standalone, diagnosis | Apr 9 |
| 07_TIMELINE.md | 14-month chronicle (Phase 1-9) + April 6-9 community acceleration + Anthropic response | Apr 9 |
| 08_UPDATE-LOG.md | Daily investigation log + changelog cross-reference | Apr 16 |
| 10_ISSUES.md | 91+ tracked issues + community tools + contributors | Apr 9 |
| 13_PROXY-DATA.md | Full proxy dataset (45,884 requests, 320 sessions, April 1–22) with Mermaid visualizations | Apr 22 |
| 14_DATA-SOURCES.md | Data label matrix (ubuntu-1-stock / ubuntu-1-override / win-1-stock), reconciliation with earlier "single machine" figures, and internal database schema overview |
Apr 16 |
| 15_ENV-BREAKDOWN.md | Per-environment cache_read ratios (pre/post April 10, daily trend), Max 20x vs Max 5x model dispatch comparison, tier-dependent Haiku share finding | Apr 16 |
| 02_RATELIMIT-HEADERS.md | Dual 5h/7d window architecture, per-1% cost, thinking token blind spot, fallback-percentage extended data | Apr 22 |
| 03_JSONL-ANALYSIS.md | Session log analysis: PRELIM inflation, subagent costs, lifecycle curve, proxy cross-validation | Apr 6 |
| 05_MICROCOMPACT.md | Deep dive: silent context stripping (Bug 4) + tool result budget (Bug 5) | Apr 15 |
| 04_BENCHMARK.md | npm vs standalone benchmark with raw per-request data | Apr 3 |
| 06_TEST-RESULTS-0403.md | April 3 integrated test results — all bugs verified | Apr 3 |
| 11_USAGE-GUIDE.md | Essential usage guide — sessions, context, CLAUDE.md, token-saving | Apr 8 |
| 12_ADVANCED-GUIDE.md | Power user guide — hooks, subagents, monitoring, rate limit tactics | Apr 8 |
- Primary dataset (this repo's published analysis):
ubuntu-1-stock- Plan: Max 20 ($200/mo)
- OS: Linux (Ubuntu), Linux workstation (ubuntu-1)
- CC mode: native
~/.claude(CC stock, no flag overrides or other instrumentation) - Versions tested: v2.1.91 (benchmark), v2.1.90, v2.1.89, v2.1.68. Changelog verified through v2.1.112
- Monitoring: cc-relay v2 transparent proxy — 45,884 total requests across 320 sessions (April 1–22)
- Parallel datasets (tracked separately, see 14_DATA-SOURCES.md):
ubuntu-1-override(same machine/account, isolated override environment with a GrowthBook flag override active since April 10 — additional components kept private),win-1-stock(Windows 11, Max 5x — research/validation only, not used for the main published analysis) - Date: April 24, 2026
This analysis builds on work by many community members who independently investigated and measured these issues. Full details in 10_ISSUES.md.
| Who | Key Contribution |
|---|---|
| @Sn3th | Discovered microcompact mechanisms (Bug 4), GrowthBook flags, budget pipeline (Bug 5) |
| @rwp65 | Discovered client-side false rate limiter (Bug 3) |
| @cnighswonger | Built cache-fix interceptor (154+ stars), 101K-call cross-validation (Jan–Apr 2026, Max 5x) with quota multipliers up to 38.6x, January baseline (unique pre-Feb data), 4-session workload comparison, DISABLE_ADAPTIVE_THINKING 3.3x reduction discovery, Explore subagent = Haiku finding (Issue #3, Issue #4) |
| @wpank | 47,810 requests tracked, v2.1.63 vs v2.1.96 quantitative comparison |
| @seanGSISG | Independent corroboration with 215K API calls (Dec 2025 – Apr 2026, Max 20x) — counterfactual proof (0 vs 18 days over budget), 6 reproducible analysis scripts (v2), iterations correlation (30.9pp), self-correcting 4.7 analysis (Issue #3) |
| @fgrosswig | 64x budget reduction forensics — 18-day JSONL analysis |
| @Commandershadow9 | 34-143x capacity reduction analysis, thinking token hypothesis |
| @kolkov | Built ccdiag, identified v2.1.91 resume regressions |
| @simpolism | Resume cache fix patch (99.7-99.9% hit) |
| @bilby91 | Identified skill_listing + companion_intro cache miss on resume |
| @labzink | Identified SendMessage cache full miss (Bug 2a) |
| @wjordan | Found "Output efficiency" system prompt change via Piebald-AI |
| @EmpireJones | Discovered telemetry-cache TTL coupling (Anthropic has repro) |
| @dancinlife | organizationUuid quota pooling discovery |
| @luongnv89 | Cache TTL analysis, built CUStats |
| @weilhalt | Built BudMon for rate-limit monitoring |
| @arizonawayfarer | GrowthBook flag dumps, acompact tool duplication analysis (35%) |
| @progerzua | /branch context inflation measurement (Bug 9) |
| Reddit community | Reverse engineering of cache sentinel mechanism |
This analysis is based on community research and personal measurement. It is not endorsed by Anthropic. All workarounds use only official tools and documented features.
