--- title: OpenAI expands Codex with Sites, Annotations, and... description: OpenAI Codex has evolved from a coding tool into a broader work platform with new capabilities including Sites (hosting web apps from natural language),... canonical: https://daily.dev/posts/openai-expands-codex-with-sites-annotations-and-role-specific-plugins-for-enterprise-workers-gfvkc6ilu twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: OpenAI expands Codex with Sites, Annotations, and role-specific plugins for enterprise workers | daily.dev og:description: OpenAI Codex has evolved from a coding tool into a broader work platform with new capabilities including Sites (hosting web apps from natural language),... og:url: https://daily.dev/posts/openai-expands-codex-with-sites-annotations-and-role-specific-plugins-for-enterprise-workers-gfvkc6ilu og:image: https://api.daily.dev/og/posts/GfVkc6ILu.png og:image:alt: OpenAI expands Codex with Sites, Annotations, and role-specific plugins for enterprise workers og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # OpenAI expands Codex with Sites, Annotations, and role-specific plugins for enterprise workers **[Collections](https://daily.dev/sources/collections)** · 5 min read · 2 upvotes · 1 comments ## Summary OpenAI Codex has evolved from a coding tool into a broader work platform with new capabilities including Sites (hosting web apps from natural language), role-specific plugins connecting to 62 business applications (Snowflake, Figma, Salesforce), Annotations for non-developers, and rate limit banking. Non-developer adoption is growing 3x faster than engineer adoption, with 20% of 5M weekly users being non-developers. The piece examines the emerging 'loop engineering' paradigm where developers design automated agent workflows rather than manually prompting. A critical analysis highlights a verification gap: while AI tools drive ~4x raw output increases, real productivity gains are only ~12%, with code churn up 861% and defect rates jumping from 9% to 54%. The bottleneck in software development has shifted from writing to verification, and teams that invest in review processes are the ones capturing real productivity gains. ## Content ## What Codex actually is now Codex started as a coding tool. It's not really that anymore — or at least, that's only part of what it does. OpenAI has been quietly expanding it into something closer to a general-purpose work platform, and the pace of that expansion has picked up noticeably. The headline numbers: 5 million weekly users, with non-developers now making up 20% of that base and growing three times faster than engineers. Internal OpenAI data claims 98% of employees use it — though every metric in that paper is self-reported by a company with obvious incentive to promote the product ahead of a potential IPO, so take that with appropriate skepticism. ## The plugin expansion The biggest structural change is the plugin system. Codex now connects to 62 business applications with 110 automated skills, organized around specific job roles rather than individual tools: - **Data Analytics** — query, report, visualize - **Creative Production** — briefs, directions, assets - **Sales** — account prioritization, buyer research, outreach drafting - **Product Design** — prototypes, specs - **Public Equity Investing** — research workflows Clay is now available as a plugin. So is Snowflake, Figma, Salesforce, and a growing list of others. The DigitalOcean plugin lets you spin up a persistent cloud development environment in a single prompt — it runs in your DigitalOcean account and keeps working after you close your laptop. For iOS developers, there's a Build iOS Apps plugin that runs an in-app simulator so Codex can view and test your app directly. OpenAI is also working on Microsoft Office integration — Excel and PowerPoint add-ins that give Codex additional control via computer use. A Record & Replay plugin captures your actions and converts them into reusable executable skills. Not available in EEA, UK, or Switzerland yet. ## Sites and Annotations Two other new capabilities worth knowing about: **Sites** lets Codex turn a description into a hosted interactive web app with a shareable URL. It's rolling out to Business and Enterprise plans first. The Codex PM described using it to turn project context into live internal tools — dashboards, focus apps, reference pages — without writing any frontend code. **Annotations** lets you give feedback directly on documents, slides, and sheets through visual markup rather than text descriptions. The agent acts on the annotations in place. ## Automations: the feature most people missed The Automations tab is probably the least-discussed part of Codex. You describe a workflow in plain language, and Codex sets it up as a recurring automated process. A practical example: monitor a GitHub repo for issues labeled "ready for agent," pick up unclaimed ones, open pull requests, and use computer use to interact with UI where needed. A built-in memory feature lets each run document its work so subsequent runs don't duplicate effort. This connects to a broader pattern several developers have been writing about. Boris Cherny, who leads Claude Code at Anthropic, mentioned he no longer prompts Claude directly — he writes loops. The idea is that most agent runs should eventually be triggered by agent-generated prompts rather than human ones. You set up the loop; the agents handle the rest. One developer described running Codex continuously for four days across multiple worktrees, with the tool building, testing, signing up for services via browser use, and filing PRs — while the human's main job was adding credit card details and closing things that weren't a fit. Another noted that top 1% Codex users average 71 agent-hours per day in a 24-hour period. The 5-hour session limit has been removed. Usage resets can now be banked and shared via referrals. ## The role shift question The expansion raises a real question about what engineering work looks like when code generation is cheap and fast. Several people have been circling the same answer from different directions: - Anthropic says 80% of its production code in May was written by Claude - GitLab research finds AI accelerates coding but not overall software delivery - Data from Faros AI, CodeRabbit, and GitClear shows roughly 4x raw output but only ~12% real productivity gain, with code churn up 861% and defect rates jumping from 9% to 54% The bottleneck has moved from writing to verification. Code review is now the highest-leverage skill in software, not because it was unimportant before, but because the volume of code requiring review has outpaced human capacity to read it carefully. The practical response most teams are landing on: tier review effort by blast radius rather than treating every PR equally, require PRs to arrive with stated intent and test evidence, run multiple heterogeneous AI reviewers rather than one, and keep a human accountable for every merge on load-bearing paths. Some developers are using Codex and Claude Code to triage PR queues — letting the agent flag high-risk changes so human attention goes where it matters rather than being spread evenly across every diff. ## What's still unclear The positioning of ChatGPT Work versus Codex has confused people. One observer described Work as "a dumbed-down version of Codex that secretly does coding but hides it" — with no clear vision for what it becomes and no new features at launch. OpenAI hasn't resolved this publicly. The self-reported adoption metrics are a real problem for anyone trying to evaluate the platform objectively. The gap between OpenAI's claimed near-universal internal adoption and external adoption rates under 1% is large enough to warrant skepticism about how representative the internal data is. And the token cost question isn't going away. Autonomous agents use roughly 30x more tokens than regular chat to complete agentic loops. At scale, that's a real budget line — Uber reportedly capped per-engineer agentic spend at $1,500/month. The economics of running loops unattended are still being worked out. ## Appshots Still one of the best features. Still underrated. ## Community discussion Top comments from developers on daily.dev. **@idoshamun** · 0 upvotes > Ok, Now that's something! Sites sounds cool, wanna try it out --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#openai](https://daily.dev/tags/openai), [#codex](https://daily.dev/tags/codex) [View this post on daily.dev](https://daily.dev/posts/openai-expands-codex-with-sites-annotations-and-role-specific-plugins-for-enterprise-workers-gfvkc6ilu) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"OpenAI expands Codex with Sites, Annotations, and role-specific plugins for enterprise workers","url":"https://daily.dev/posts/openai-expands-codex-with-sites-annotations-and-role-specific-plugins-for-enterprise-workers-gfvkc6ilu","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/openai-expands-codex-with-sites-annotations-and-role-specific-plugins-for-enterprise-workers-gfvkc6ilu"},"datePublished":"2026-06-02T16:27:24.694Z","dateModified":"2026-07-13T21:02:10.681Z","description":"OpenAI Codex has evolved from a coding tool into a broader work platform with new capabilities including Sites (hosting web apps from natural language),...","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/openai-expands-codex-with-sites-annotations-and-role-specific-plugins-for-enterprise-workers-gfvkc6ilu","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"ai-agents,openai,codex","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"OpenAI expands Codex with Sites, Annotations, and role-specific plugins for enterprise workers"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/openai-expands-codex-with-sites-annotations-and-role-specific-plugins-for-enterprise-workers-gfvkc6ilu","comment":[{"@type":"Comment","text":"Ok, Now that’s something! Sites sounds cool, wanna try it out","datePublished":"2026-06-02T18:17:49.971Z","url":"https://daily.dev/posts/GfVkc6ILu#c-CIpTDuEuS","author":{"@type":"Person","name":"Ido Shamun","url":"https://daily.dev/idoshamun","image":"https://media.daily.dev/image/upload/s---xy_OAwk--/f_auto,q_auto/v1703781380/avatars/avatar_28849d86070e4c099c877ab6837c61f0"}}]} ```