Frontier Inference Margins
An honesty-first interactive cost model of frontier-LLM inference serving margins — mechanism-first calculator, typed public-claims registry, and a long-form research report. Every number carries its epistemic status.
- Mechanism-first calculator — a ≈77% central-scenario modeled unit serving margin (not a company gross margin), every assumption adjustable
- 34-claim typed evidence registry with provenance tiers
- Read-only MCP connector at margins-mcp.ashitaorbis.com
- Three test suites and published release gates behind every deploy
Activity Timeline
-
CI hardened; GPT-Pro daily backlog through 2026-07-25 committed.
Read-only token, checksum-verified shfmt, and .pyc cleanup applied to CI. Daily material backlog for prior week persisted.
-
TPU7 +495.4% miss traced to operating-point-closure error; research pipeline automated.
Root cause: design inverted b*=154/chip vs actual ~16 concurrent-sequence ceiling. Standing recommendation: anchor future designs to 64-sequence batch ceiling. FlexNPU CM384 extracted for cycle-2. ChatGPT Pro polling now fully automated — dispatch, poll, persist.
-
Cycle 2 root cause found: 16 seq/chip actual vs ~154/chip assumed, explains +495.4% miss.
Operating-point-closure error traced and verified (TPU7 Ironwood: 64 concurrent sequences, 16/chip, 518.86 tok/s/chip). CM384 FlexNPU extracted via parallel research agents. Epoch AI and wafer_ai added to hardware sweep; GPT Pro replica-width consult refuted both prior candidate bounds.
-
IM3 exit verification in final rounds; IM4 diagnostic harness spec locked.
P0/P1 defect classes fixed and re-verified through Rounds 2–3. Script defect recovery advanced from r7 to r12 (r11 released defect-free). IM4 entry harness specified with survivorship regression fixtures.
-
ChatGPT Pro MCP disk_recovery premature-completion bug fixed; daily sweep on schedule.
Root cause of premature completion in disk_recovery path identified, documented, and patched in one commit. Daily automation pipeline persisting 2026-07-19 research findings (Z.ai, DeepSeek/Kimi, Anthropic credit line) to gptpro-reports/.
-
v2.2 verification in progress; deploy awaiting owner approval.
Ground-truth reconciliation pass underway against effective-MFU scalar model. ChatGPT Pro MCP completion-detection bug filed (commit 53b3dac). Deployment staged pending sign-off.
-
Pipeline hardened (4 P1 + 1 P2 findings applied); v2.2 on HOLD pending owner sign-off.
D2 receipt pack completed with full model/platform/operating-point verification. Cycle-2 re-engineering underway but threshold finalization blocked on owner gate.
-
v2.1.6→v2.1.11 (6 releases, 44 commits): billing bug fixed, GPT-Pro wired, cold review hygiene pass.
Chart billing canonicalized to shared computeMix engine with regression tests. Four dive reports scrubbed of provenance headers, analysis preserved. GPT-Pro daily/weekly reports live via bash-owned fetcher with in-turn polling and 70-min codex fallback.
-
v2.1.4: evidence-scent label pass, outside-review remediation, feedback intake live.
Page-set values relabeled so they no longer carry a false empirical scent; GPT-Pro outside-review findings remediated and the engine bumped to v2.1.4; quarantined feedback intake (Cloudflare Worker + D1 + Turnstile) embedded on the main page.
-
Margin-range evidence board redesign shipped; MCP connector deployed (v2.1.3).
Preset dropdowns replaced by a mechanism-first margin-range evidence board with de-named routes and v4 permalinks. Read-only MCP server deployed as a Cloudflare Worker at margins-mcp.ashitaorbis.com behind a fixture-gated honesty envelope.
-
Traffic-mix axis added to the engine (v2.1.2); reception audit round remediated.
New traffic-mix axis with a pure codec and atomic permalink replays, specified first by failing-by-design contract fixtures. Nine reception audits run against the live site; all 16 P0 findings remediated and the resolution ledger published.
-
Initial build and first production deploy.
Honesty-first interactive cost model of frontier-LLM inference serving margins: mechanism-first calculator, typed public-claims registry, and a long-form research report, deployed to Cloudflare Pages.