OpenAI Research — ChatGPT Desktop (Modos ChatGPT e Codex)
Hub da pesquisa de engenharia reversa do ecossistema OpenAI — protocolos, probes, vendors e estado da arte
BM25 9.37 · title +1.50 · PageRank +0.70pesquisaChatGPT Desktop — Hybrid Architecture and Mode Hierarchy
RE of the desktop app — Swift shell + Chromium renderer + embedded Codex app-server, the three conversation surfaces and their transports
BM25 9.50 · title +1.50 · PageRank +0.18pesquisaFlow A Manual — Step-by-Step (executable by a fresh agent)
The validated end-to-end sequence of flow A1 (session-keeper) with exact headers, bodies and the live token acquisition — reproducible request by request via the bezetacil harness
BM25 8.89 · PageRank +0.35pesquisaFlow Definitions — A / A0 / A1 / B (attestation vs PoW)
The definitive hierarchy of the ChatGPT Desktop flows — attestation (A, with A0 standalone and A1 session-keeper variants) and the untested sentinel/PoW path (B)
BM25 8.69 · PageRank +0.38pesquisaBezetacil v6 — Runtime Harness for Electron Apps
Injector/manager with hooks + on-demand triggers, REST server, per-session log rotation, git-versioned home — deployed on the ChatGPT desktop
BM25 8.61 · PageRank +0.22pesquisaAnálise de Vendors — OmniRoute, openai-oauth e pi-ai
What each open source solution gets right, the errors that cause bans, and why none knows the critical mechanisms
BM25 8.54 · PageRank +0.12pesquisaAttestation Flow — DeviceCheck, Challenge, e o Hierarchy com PoW Fallback
The complete integrity mechanism of the desktop app — DeviceCheck native addon, challenge flow, and the if/else hierarchy with PoW fallback for non-Apple platforms
BM25 8.41 · PageRank +0.27pesquisa001-net-fetch-hook — electron.net.fetch Hook
Bezetacil script that intercepts every HTTP request of the main process with stack capture
BM25 8.49 · PageRank +0.07pesquisaCodex Mode Runtime Trace — Mode Inside the Desktop App
Live evidence of the Codex mode (top-level, no submodes) inside the ChatGPT desktop — webrun/alpha-search tool, model switch, task-title agent, and the differences vs the work submode
BM25 8.40 · PageRank +0.06pesquisaAuth, Preemptive Refresh, and Link-Session
Self-contained OAuth PKCE, token rotation with preemptive refresh (codex-rs semantics), link-session exchange, and cross-surface Bearer
BM25 8.29 · PageRank +0.19pesquisaSentinel Fallback Validation — Exercising the PoW Path on macOS
Runtime validation of the sentinel/PoW fallback by flipping the DeviceCheck gate — real fingerprint values, real headers, placeholder confirmation
BM25 7.75 · PageRank +0.22pesquisaVoz e Tasks no ChatGPT Desktop
Ditado (/transcribe), live voice thread-scoped (WebRTC realtime) e tasks do Work criadas pelo Chat (/wham)
BM25 6.94 · title +1.50 · PageRank +0.13pesquisaFlow A Standalone — Verified Building Blocks
Every fact needed to run the attested chat-submode flow outside the desktop app — refresh (oai_is), Node DeviceCheck token, real headers/bodies, conduits, attestation challenge
BM25 7.72 · PageRank +0.05pesquisaCodex Backend (App-Server) — Responses, Sticky Routing, Server-Side Compact
The codex backend protocol of the embedded app-server — official endpoints, encrypted reasoning, quota snapshot, and the Rust HTTP surface
BM25 7.51 · PageRank +0.23pesquisaSentinel Fingerprint Comparison — Desktop vs OmniRoute
The PoW/requirements fingerprint arrays field-by-field — what each client sends, what the server validates, and the probe corrections
BM25 7.39 · PageRank +0.14pesquisaModo ChatGPT, Submodo Chat — Protocolo Real (traces da sessão)
The chat submode protocol reconstructed from 40 captured traces — typing throttle, attestation, UI efforts, multi-turn
BM25 7.01 · PageRank +0.39pesquisaToken Codex — Claims, Scopes, and JWT Structure
The full JWT decoded from the Bearer token used by the desktop app — claims, scopes, and what each field enables
BM25 6.86 · PageRank +0.06pesquisaWork Mode Runtime Trace — The Full Execution Loop
Work Mode Runtime Trace — The Full Execution Loop All facts below come from a live session 2026 08 26 : a task created in the chat submode submodo chat handed off to work mode modo work run once GPT 5.5 then corrected GPT 5.6 with model/...
BM25 6.59 · PageRank +0.08outroWikifita — Log
Wikifita — Log Chronological log of wiki operations. 2026 09 04 fix CAMDOM Mikael/Maso Heck identity reconciliation Corrected the CAMDOM people graph so Mika Mikael and Mikaelzinho consistently identify Mikael Alefita's husband and techn...
BM25 5.68 · PageRank +0.03pesquisaModo ChatGPT, Submodo Work — Protocolo
As duas formas de iniciar o submodo work (direto pela aba e via handoff), execution target local_executor, e o que muda no payload
BM25 5.44 · PageRank +0.15pesquisaProof of Work — Sentinel PoW Mechanism (Desktop vs Web)
Proof of Work — Sentinel PoW Mechanism Desktop vs Web Part of openai research. The sentinel PoW is the integrity mechanism used by the web/browser route and by the desktop fallback when DeviceCheck is unavailable — the gate flip . What i...
BM25 4.00 · PageRank +0.08pesquisaUser Agents — Every UA the Desktop App and Its Clients Send
User Agents — Every UA the Desktop App and Its Clients Send Part of openai research. A standalone must present the right User Agent per surface — servers route/validate by client type this is also hypothesis for detections . Captured UAs...
BM25 3.97 · PageRank +0.05projetoClaude Desktop — Feature Flags
Documentation: Claude Desktop — Feature Flags
BM25 3.80 · PageRank +0.29projetoClaude Desktop — Bundle Architecture
Documentation: Claude Desktop — Bundle Architecture
BM25 3.74 · PageRank +0.33