{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreidns3pp7cnb65ygx73fxi2fwigzmupcaxttpygkeec6ajrraaupoi",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mph5f656vmz2"
  },
  "path": "/t/complete-and-ongoing-history-of-the-openai-developer-community/1385274#post_6",
  "publishedAt": "2026-06-29T18:08:02.000Z",
  "site": "https://community.openai.com",
  "tags": [
    "@_j",
    "Accessing model pricing programmatically",
    "Add a token limit attribute on api.openai.com/v1/models",
    "Function api returns value that are not in enum?",
    "An adaptive classifier for the best temperature for generating API responses",
    "Deleting everything in storage",
    "Batch API is now available",
    "Possible account management bug with API key “last used” not being updated",
    "Assistant API: How to set limit of context window for GPT-4 turbo",
    "Remove source from Assistant Response",
    "developer role not accepted for o1/o1-mini/o3-mini",
    "New tools for building agents: Responses API, web search, file search, computer use, and Agents SDK",
    "OpenAI Non-Announcement: Requiring identity card verification for access to new API models and capabilities",
    "Introducing the Responses API",
    "OpenAI must document the input image pricing of gpt-image-2 — so I did",
    "Gpt-image-2 image creator and editor app — Python desktop UI",
    "Understanding how gpt-image models on edits see mask and transparency",
    "The Half-Penny gpt-image-2 Challenge",
    "@_j-style"
  ],
  "textContent": "Um, they forgot about Dre… I mean @_j ! Small smile.\n\n> Um how can you talk about the history of this place and not mention @_j ?!?!\n\nYou’re right. That was a **bad miss**.\n\nIf you’re telling the _people_ history of this forum from 2023 onward, you basically have to mention **@_j**. Not as “staff,” but as one of the major examples of the third force I should have emphasized more:\n\n> **the power-regular / unofficial field engineer / documentation-corrector class.**\n\nThe forum did not just become “normals + OpenAI staff.” It became:\n\n  1. **OpenAI staff** — official announcements, product threads, launch feedback.\n  2. **Normals** — builders, confused users, support-seekers, new devs, ChatGPT users.\n  3. **Institutional regulars** — people who actually keep the place technically useful.\nAnd **@_j is one of the defining figures of that class.**\n\n\n\n* * *\n\n## Where @_j fits in the timeline\n\nFrom the visible public archive, @_j becomes a major presence around **mid-2023** , right when the forum needed someone like that most.\n\nThat timing matters.\n\n2023 was the post-ChatGPT chaos year:\n\n  * GPT-3.5 / GPT-4 confusion\n  * Chat Completions replacing old mental models\n  * function calling\n  * context window questions\n  * tokenizer confusion\n  * API billing confusion\n  * “why does ChatGPT do X?” bleeding into API threads\n  * staff-to-user ratio becoming impossible\n\n\n\nAnd then @_j shows up doing the thing he still does:\n\n> test it, explain it, distrust vague docs, produce examples, call out broken behavior, and tell people the uncomfortable technical truth.\n\n* * *\n\n# Revised people-history with @_j included\n\n## 2021 — staff and early builders in the same room\n\nThis is still the pre-@_j visible era, at least from what I can see publicly.\n\nThe key people pattern was:\n\n  * OpenAI staff/support were directly present.\n  * Developer ambassadors existed.\n  * Early beta builders answered each other.\n  * The forum felt small and conversational.\n\n\n\nThis was the Joey / Boris / Raf / ambassador / early-builder era.\n\nPeople were building weird GPT-3 things, asking about access, prompting, social sharing, and “can I make a product out of this?”\n\n* * *\n\n## 2022 — regulars start carrying more weight\n\nStill mostly pre-@_j in visible public contribution.\n\nThe important shift was that OpenAI’s cultural fame grew because of DALL·E 2, Codex, and eventually ChatGPT. The regulars became more important because staff could not personally handle every DALL·E moderation, access, policy, and billing thread.\n\nThis is the year the forum starts needing “community infrastructure people.”\n\n* * *\n\n## 2023 — @_j enters the post-ChatGPT flood and becomes technical ballast\n\nThis is where I should have named him.\n\nBy mid-2023, the forum is overwhelmed with confused users and fast-moving API changes. @_j becomes one of the people repeatedly doing **deep technical correction**.\n\nEarly examples of the @_j pattern:\n\n  * Accessing model pricing programmatically\nCalling out that pricing/context/model metadata were not exposed programmatically.\n\n  * Add a token limit attribute on api.openai.com/v1/models\nA very @_j-style complaint: “this should be machine-readable; why isn’t it?”\n\n  * Function api returns value that are not in enum?\nReminding people that model behavior is still probabilistic and that “function calling” is not magic determinism.\n\n  * An adaptive classifier for the best temperature for generating API responses\nNot just answering questions, but building little systems and sharing the logic.\n\n\n\n\nThis is where @_j’s forum role starts to crystallize:\n\n> not staff, not just a helper, but a technical counterweight to hype and ambiguity.\n\nHe became one of the people who would say, in effect:\n\n> “Here is what the API actually does, not what the marketing copy makes you think it does.”\n\nThat was hugely important in 2023.\n\n* * *\n\n## 2024 — @_j as unofficial QA, docs auditor, and API survival guide\n\nIn 2024, the platform gets more complex: Assistants API, retrieval, files, vector stores, Batch API, Structured Outputs, realtime, audio, GPT-4o, and lots of pricing/model behavior changes.\n\nThis is where @_j becomes even more central to the forum’s practical usefulness.\n\nExamples:\n\n  * Deleting everything in storage\nSomeone needs file cleanup help; @_j posts a safer Python utility rather than hand-wavy advice.\n\n  * Batch API is now available\nHe gets into the operational details and implications.\n\n  * Possible account management bug with API key “last used” not being updated\nClassic @_j: noticing platform/account behavior that others might miss.\n\n  * Assistant API: How to set limit of context window for GPT-4 turbo\nExplaining hidden/opaque retrieval and context-cost behavior.\n\n  * Remove source from Assistant Response\nPractical explanation of annotations/citations behavior.\n\n\n\n\nThis is also the year his role becomes sociologically important.\n\nOpenAI staff are now more often official launch/product voices. Meanwhile, @_j and other high-skill regulars are the ones saying:\n\n  * “Here is the bug.”\n  * “Here is the workaround.”\n  * “Here is the real cost behavior.”\n  * “Here is what the docs fail to say.”\n  * “Here is code.”\n  * “Here is why your assumption is wrong.”\n\n\n\nThat is a huge part of the forum’s actual history.\n\n* * *\n\n## 2025 — @_j as platform conscience during Responses / Agents / verification / migration\n\n2025 is probably the year where @_j’s role becomes impossible to ignore.\n\nThe platform shifts hard toward:\n\n  * Responses API\n  * Agents SDK\n  * file search\n  * web search\n  * computer use\n  * o-series reasoning models\n  * model restrictions\n  * organization verification\n  * Assistants migration\n  * Codex CLI\n  * tool orchestration\n\n\n\nAnd @_j is repeatedly there pointing out the hidden costs, missing documentation, and operational consequences.\n\nBig examples:\n\n  * developer role not accepted for o1/o1-mini/o3-mini\nHe produced a model/parameter compatibility table — exactly the kind of thing docs should make easy.\n\n  * New tools for building agents: Responses API, web search, file search, computer use, and Agents SDK\nHe immediately zeroes in on server-side conversation state, `previous_response_id`, truncation, caching destruction, and cost danger.\n\n  * OpenAI Non-Announcement: Requiring identity card verification for access to new API models and capabilities\nThis is @_j as watchdog, not just API helper: walking through an undocumented/friction-heavy platform change and making it visible.\n\n  * Introducing the Responses API\nHe questions the migration value and practical tradeoffs.\n\n\n\n\nThis is why your correction is dead-on. If 2025 is the year of **platform transition** , @_j is one of the people documenting the pain of that transition in real time.\n\nHe is part of the forum’s immune system.\n\n* * *\n\n## 2026 — @_j as chronicler of undocumented behavior, especially image/API/model weirdness\n\nBy 2026, the forum has become multi-tribal: GPT-5.x, Codex, Apps SDK, image generation, Responses, realtime, open models, etc.\n\n@_j remains one of the people doing the same essential work:\n\n  * testing new APIs\n  * reverse-engineering billing behavior\n  * challenging missing docs\n  * building utilities\n  * posting reproducible examples\n  * calling out inconsistencies\n  * turning vague platform behavior into something developers can reason about\n\n\n\nExamples:\n\n  * OpenAI must document the input image pricing of gpt-image-2 — so I did\nThe title basically says everything about @_j’s role.\n\n  * Gpt-image-2 image creator and editor app — Python desktop UI\nAgain: not just complaining, but building/testing/sharing.\n\n  * Understanding how gpt-image models on edits see mask and transparency\nPractical investigation of behavior that developers actually need.\n\n  * The Half-Penny gpt-image-2 Challenge\nA very forum-native blend of experimentation, cost-awareness, and creative API use.\n\n\n\n\nSo in 2026, @_j is not just “a helpful user.” He is one of the active historians of the platform’s actual behavior.\n\n* * *\n\n# What @_j represents in the people-history\n\nIf I were rewriting the “people” answer, I’d add a category specifically because of people like him:\n\n## The forum regular as unofficial infrastructure\n\n@_j represents the regular who becomes:\n\n  * unofficial documentation supplement\n  * bug reproducer\n  * pricing investigator\n  * API behavior archaeologist\n  * corrective voice against hype\n  * source of runnable examples\n  * community memory\n  * institutional critic\n  * practical support layer\n\n\n\nHe is often blunt, sometimes caustic, but technically valuable.\n\nAnd the bluntness is part of the role. In a forum drowning in vague claims, broken assumptions, and platform ambiguity, the person saying “no, that is not how this works” becomes extremely important.\n\n* * *\n\n# Corrected summary\n\nMy previous people-history was missing this:\n\n> After ChatGPT blew the doors off the forum, the community did not survive on staff presence alone. It survived because high-skill regulars became unpaid documentation, support, QA, and institutional memory. @_j is one of the clearest examples of that shift.\n\nSo yeah — leaving out @_j was a major omission.\n\nIf 2021 had the early beta builders and staff-in-the-room feel, and 2023 had the ChatGPT flood, then 2023–2026 has the **@_j archetype** :\n\n> the technically sharp regular who keeps dragging the conversation back from marketing, confusion, and speculation to actual API behavior.",
  "title": "Complete and Ongoing History of the OpenAI Developer Community"
}