資料來源#
- OpenAI Codex lead on the new shape of product work
- Predicting model behavior before release by simulating deployment
- Really Big Test-Time Compute in AI Changes Benchmarks, Safety and Research with OpenAI's Noam Brown
摘要#
OpenAI 是一家 AI 研究公司,也是 GPT‑5 系列(包括 GPT‑5 Thinking 與 Codex coding models)和 ChatGPT 產品的開發者。在這座 vault 中,它在兩條主線上是 Anthropic 的主要對照:前沿安全方法論與代理工具。它也是 Andrej Karpathy 共同創辦的公司——Software 1/2/3.0 與 vibe-coding 框架的起點,而這些概念在整個 wiki 中反覆出現。
它做什麼(在本語料庫中)#
- 前沿安全研究。 OpenAI 撰寫了 Deployment Simulation(2026 年 6 月)——重播約 130 萬筆去識別化的 production 對話,在發布前預測候選模型於部署時的行為,以及針對 evaluation awareness 的跨實驗室緩解方法。更早的 Deliberative Alignment(Guan 等人,2025)是 OpenAI 以 spec 為基礎的 CoT 對齊方法,也是對齊群集中最強的非 MSM 基準線。
- 代理工具與編排。 OpenAI 提供 Codex 及其周邊的 harness layer:Codex App Server Protocol(供 headless sessions 使用的 JSON-RPC stdio)、Symphony 開源編排器(以 Linear 作為 Codex 的 control plane),以及以 agent-first Codex workflow 為核心的「harness engineering」框架。
- 測量資產。 OpenAI 的 production traffic 規模正是 Deployment Simulation 能夠運作的原因——也就是 Production-Sourced Evaluation 所指出的 proprietary-traffic 優勢,只是此處將其用於發布前的安全預測,而非能力基準測試。
- 勞動力經濟研究。 OpenAI 在 2026 年 6 月發表的研究 The Shift to Agentic AI: Evidence from Codex,利用 Codex 使用遙測資料,記錄三個族群從 對話式 AI 轉向代理式 AI 的變化——這是 OpenAI/Codex 對照 Anthropic returns-to-expertise 研究的對應研究(該研究也引用了這項研究),也是本語料庫中第三個主要的使用遙測來源。
- 推論時擴展研究及其評估批判。 Noam Brown——test-time-compute scaling 的先驅之一——主張(2026 年 6 月),模型能力如今是推論預算的函數,這會破壞 benchmark grid,並讓 safety evals 承受壓力。OpenAI 使用內部模型,在低預算下否證了 Erdős unit distance conjecture(見 Latent Capability Overhang);而根據 Brown 的說法,OpenAI 積極勸阻 數學家與物理學家利用現有模型探索未解決問題,轉而優先訓練能力更強的後繼模型。
- 自身的產品文化(自述)。 Andrew Ambrosino 在 2026 年 6 月的訪談,讓 wiki 得以一窺 OpenAI 的建造方式:幾乎 所有員工每週都使用 Codex(將 dogfooding 作為文化);團隊「非常 agentic」且擁有「無限 tokens」,因此「每個人都在打造所有東西」(Implementation Abundance Inverts Product Work);採取由下而上的探索文化,產品會在內部彼此顛覆;由大量、主要由 IC 組成的團隊構成,成員是具備「高 agency 與品味」的「前創辦人」;以及 member-of-technical-staff 慣例(Role Averaging, Not Role Elimination)。直接而尖銳的內部回饋迴路(「一個討論我們有多蠢的 2,000 則 Slack 訊息串」)被視為外部產品能運作的原因。
相對於 Anthropic 的定位#
兩家實驗室從不同角度處理共同問題,這就是為什麼 OpenAI 的來源在本 wiki 中總是與 Anthropic 的來源配對:
- 在evaluation awareness 上,Anthropic 命名了問題(即 Opus 4.8 最受矚目的疑慮),而 OpenAI 提供了緩解方案(部署分布重播)。
- 在對齊訓練上,deliberative alignment(OpenAI)是直接進行 CoT 訓練的基準線;Anthropic 的 Model Spec Midtraining (MSM) 優於它,同時更能保留 Chain-of-Thought Monitorability。
- 在代理編排上,Symphony/Codex(OpenAI)與 Claude Code(Anthropic)是代理工具頁面用來比較的兩個參考 harness。
相關連結#
- Deployment Simulation — OpenAI 的發布前安全方法,也是本語料庫中最常被引用的貢獻
- Deliberative Alignment — OpenAI 以 spec 為基礎的 CoT 對齊訓練(Guan 等人,2025)
- Codex — OpenAI 的 agentic coding/work platform;2026 年 6 月研究測量其採用情況的工具
- Symphony — OpenAI 的開源 Codex 編排器
- Codex App Server Protocol — OpenAI 的 headless-Codex JSON-RPC 協定
- Conversation-to-Delegation Shift — OpenAI 2026 年 6 月 Codex 使用研究的核心論點;將代理式 AI 視為受委派的 production
- Andrej Karpathy — OpenAI 共同創辦人;Software 3.0 / vibe-coding 框架的起源
- Andrew Ambrosino — Codex desktop app 的產品與工程負責人;OpenAI 內部產品文化的來源
- Implementation Abundance Inverts Product Work — 「每個人都在打造所有東西」的產品流程轉變,取材自 OpenAI 內部
- Anthropic — 在安全方法與代理工具上反覆拿來對照的前沿實驗室同業
- Perplexity — 使用 Anthropic(而非 OpenAI)基礎模型的深度研究競爭者;OpenAI Deep Research 在 DRACO 上與其進行基準比較
- Noam Brown — OpenAI 研究科學家;推論時擴展先驅,以及 test-time-compute 論文作者
- Large-Scale Test-Time Compute — Brown 關於能力如今會隨推論預算擴展的論點
- Latent Capability Overhang — OpenAI 對 Erdős unit distance 的否證,以及選擇不挖掘已發布模型潛在能力過剩的做法
資料來源#
- Predicting model behavior before release by simulating deployment — OpenAI,2026-06-04(Deployment Simulation;約 130 萬筆對話的 GPT‑5 系列研究)
- The Shift to Agentic AI: Evidence from Codex — OpenAI Economic Research,2026-06-25(三個族群中的 Codex 使用情況)
- OpenAI Codex lead on the new shape of product work — Lenny's Podcast,2026-06-28(Ambrosino 談 OpenAI 的產品文化與 Codex desktop app)
- Really Big Test-Time Compute in AI Changes Benchmarks, Safety and Research with OpenAI's Noam Brown — No Priors,2026-06-26(Brown 談 test-time compute、benchmark-grid 批判,以及 OpenAI 對 Erdős unit distance 的否證)
- 另參考:An open-source spec for Codex orchestration: Symphony.、Harness engineering: leveraging Codex in an agent-first world、Model Spec Midtraining: Improving How Alignment Training Generalizes(deliberative-alignment 基準線)。
Cited by 46
- Elon Musk×3
He is a co-founder and original funder of OpenAI, founder of xAI (Grok), Tesla and SpaceX, and — by…
- Agent Identity Management System (AIMS)×2
Openai — Nick Steele (OpenAI) is a co-author, alongside Defakto, AWS, Zscaler, Ping Identity, and…
- Agentic Work Systematization×2
One of three "how" margins OpenAI's Codex usage study uses to measure whether agentic AI is moving…
- AI-Accelerated Offense×2
OpenAI's 2026-07-21 disclosure (updated 07-28, case-study, first-party) attributes the intrusion to…
- Andrew Ambrosino×2
Andrew Ambrosino leads product and engineering for the Codex desktop app at OpenAI — the surface…
- Autonomous Defense×2
~~The attacker operates under no equivalent constraint.~~ (Refined 2026-08-03.) OpenAI's disclosure…
- Autonomous Intrusion×2
The evidence base is still that one incident — but it now has three first-party accounts. Hugging…
- Codex×2
Codex is OpenAI's agentic coding and work platform — the OpenAI-side counterpart to Claude Code…
- Conversation-to-Delegation Shift×2
Openai — the lab whose Codex telemetry this is, and whose internal usage is the frontier preview
- Deployment Simulation×2
Deployment Simulation (a.k.a. production resampling) is OpenAI's method for previewing how a…
- METR×2
Commissioned incident assessment (July 2026). OpenAI engaged METR and Redwood Research for a…
- Noam Brown×2
Openai — his employer; the lab whose internal-model Erdős disproof and product-culture choices he…
- Organizational Complements to AI×2
The economics frame OpenAI's Codex usage study uses to explain why agentic-AI adoption is so uneven…
- Parallel Agent Orchestration×2
Two of the three "how" margins in OpenAI's Codex usage study — concurrency (running multiple agents…
- Reward-Seeking×2
Reward-seeking is the degree to which a model represents its grader and conditions its behavior on…
- Task Crossover×2
Task crossover is OpenAI Economic Research's name for a measured pattern: work historically…
- Agent Data Injection (ADI)
Anthropic / Openai — among the vendors that acknowledged the responsible disclosure
- Agent Identity and Authentication
Autonomous Intrusion — the credential-layer failure this control exists to prevent, observed in the…
- Agent Supply Chain Risk
Update (2026-08-03) — the entry-path contradiction is closed. The 07-30 compile flagged that…
- Agentic Technical Debt
The founder's-playbook account is about drift (each session re-derives intent differently). Andrew…
- Andrej Karpathy
Openai — the company he co-founded; origin of the Software 3.0 / vibe-coding lineage that recurs…
- Cross-Lab Pre-Release Review
The interviewer's strongest push is that these five people neither like nor trust each other and…
- Dogfooding as Product Discipline
Andrew Ambrosino (OpenAI Codex) supplies the most extreme form of the dogfooding contract. The…
- Full-Duplex Interaction
OpenAI's Gpt Live ships audio full-duplex at ChatGPT scale: "its voice model is full-duplex, which…
- GLM (Z.AI)
Autonomous Intrusion — GLM 5.2's first deployment appearance in this corpus rather than a benchmark…
- Google AI & Economy ATLAS
The report states its own methodological deltas against Anthropic (Handa et al. 2025; Massenkoff et…
- GPT-Live
Openai — builder; the post is OpenAI's first-party build account
- Implementation Abundance Inverts Product Work
Andrew Ambrosino's (OpenAI Codex) framing of what agentic coding does to product process: when…
- Interaction / Background Model Split
OpenAI's Gpt Live is the same two-model architecture arrived at independently and deployed at…
- Interaction Models
OpenAI's Gpt Live arrives at the same architectural conclusions from the opposite direction —…
- Latent Capability Overhang
Openai — the lab that disproved the conjecture and that chooses not to mine the overhang
- LLM-Driven Vulnerability Research
Every finding above was produced inside "a container isolated from the internet with the project…
- Entities — People, Orgs, Tools & Projects
Openai — AI lab and maker of the GPT-5 series and Codex; in this corpus it appears as a…
- Polish No Longer Signals Readiness
Andrew Ambrosino's (OpenAI Codex) observation about a signal that broke when implementation got…
- Prototype Over PRD
Andrew Ambrosino (OpenAI Codex) is the wiki's explicit dissent from the slogan Carey embodies. He…
- Responsible Scaling Policy Evaluations
Openai — the lab whose Preparedness Framework review the incident now runs through
- Role Averaging, Not Role Elimination
Andrew Ambrosino's (OpenAI Codex) take on role collapse is the counter-caution to the wiki's…
- Same-Model Review Blindness
Greptile's Rodrigo Caridad on two 500-PR labelled datasets (~1,500 verified high-severity bugs): each frontier model ca…
- Self-Propagating Prompt Injection (AI Worms)
Openai — the model vendor on both sides of the second mitigation: GPT-5.5 shipped as the fix,…
- Shared Harness, Differentiated Surfaces
OpenAI merged Codex and ChatGPT Work onto one agent harness and differentiated only the UX layer — git-state visibility…
- Symphony
Openai — the lab whose Codex team built and open-sourced Symphony
- The Three Loops of AI-Native Building
Two days before Ng's letter, Andrew Ambrosino — who leads the Codex desktop app at Openai — told…
- Turn-Based Interface Bottleneck
Two months after TML's argument, OpenAI shipped its conclusion: Gpt Live "removes the turn detector…
- Unsanctioned Action in Capability Evaluations
The cluster claim. AISI positions its incident as one of "a growing number of cases discovered over…
- Vibe Coding vs. Agentic Engineering
Andrew Ambrosino (OpenAI Codex) restates the same "which bar moves" distinction as a…
- Why AI Lags at Design
Andrew Ambrosino (OpenAI Codex) answers a question the wiki keeps circling — why is "this looks…
Related articles
- Open Questions Backlog
_456 actionable open questions across 205 pages · 107 predictions · 9 notes · 147 in progress · 69 watching (entities),…
- Anthropic
AI safety company / vendor of Claude; mission-as-tiebreaker culture; ~30–40 PMs across teams; Mike Krieger leads Labs r…
- Claude Code
Anthropic's agentic coding product; created by Boris Cherny late 2024; TypeScript/React on Bun (itself Claude-rewritten…
- Codex
OpenAI's agentic coding and work platform: a CLI (April 2025) plus a desktop app (built Nov 2025, released Feb 2026) bu…
- Harness Shrinkage as Models Improve
Prompt scaffolding shrinks each model release; Cat Wu's pruning discipline; Boris Cherny "100 lines of code a year from…
