Alexander Embiricos

yes it is

yep!

working on it!

@ajambrosino is the new Thursday Enforcer. Thank you 🫡

@ajambrosino windows phone next right?

update the app, go to settings then computer use, and get it from there

we're coming to other chromium based browsers soon!

props to @justin_rushing, @jnpdx, @ankushg, @JamesZmSun and more. this is so good 🙌

Not yet sorry. Working on it

no support beyond Chrome, yet

Mike Isaac (NYT) Joins, Earnings Review, Beautiful Data Centers, Coatue is Land Maxxing

we fixed a top request for testing different viewport sizes in the codex in-app browser:

this checks out! good model + the basics. funnily enough i still think we're in the phase of building the basics: stability, simple UX, ability to use the same tools the user can...

Dan gets it! That said doing all your browsing in codex is certainly a **power** user move today. Planning to make that much better.

🫡

Thinking about all of this! Would love to hear more about how you use that. Open to showing me over a call?

Little Friday bonus for the Codex app: Pets, a delightful way to keep track of Codex working in the background. Wake your pet with `/pet`

thanks! we had quite some debate before shipping pets. we were worried that folks might pan us for not being 100% focused on raw capability. but hopefully 99% capability and 1% fun works for you all!

CODEX_HOME=~/.codex-onboarding-test-4 pnpm run electron

upgrade your agent without starting from scratch

+1. There's also some product building strategy in here: we're setting things up so that all the capabilities snowball into a single implementation. e.g. if software engineers get subagents or `/goal`, so does every other user.

BTW /goal just shipped to the CLI. When it reaches the app, it'll be for everyone. https://t.co/IOGFve27LX

Most significant improvement we shipped today: Codex asks you what work you do, and customizes itself to you.

Thanks a lot! 1/ agree 2/ looking at next week for a first version 3/ interesting! have examples? 4/ yes, exist 5/ coming soon

thanks! will reach out

fun! what changed?

throwback to founder days

codex gets faster

Codex limits! (Actually all our agents use the same limit rate pool.)

Doing our best!

Embiricos Offers Direct Assistance for User ID or Email Issues on X

Alexander Embiricos is conducting an hourly poll on his X feed and responding to a user issue by requesting their user ID or email. This enables the team to investigate the problem promptly. Indicates active user support protocol in social media operations.

OpenAI Codex Trained with Embiricos Team Assistance

OpenAI's Codex model was developed using a specific methodology outlined in their "Unlocking the Codex" announcement. Embiricos confirms direct team involvement in Codex's training process. This collaboration highlights external contributions to foundational AI model development.

Embiricos Seeks User Setup Details for Reported Issue

Alexander Embiricos acknowledges a user's report on his X feed and requests additional information about their setup via reply or DM. This interaction occurs within an hourly poll context on his feed. It demonstrates proactive engagement with user feedback.

Embiricos Takes Poll Report Seriously Despite Doubt

Alexander Embiricos expresses skepticism about a report on his X feed's hourly poll but commits to investigating it thoroughly. The response underscores a policy of treating user reports with high priority regardless of initial belief. This reflects standard platform moderation practices on X.

that shouldn't have changed. send a screenshot of what you're seeing?

@_simonsmith thanks for this. can you share /feedback for this conversation? also, what happens when you try the same prompt in with Auto-review or Full Access?

Embiricos Launches Hourly Poll on Codex's Deep Research Query Performance

Alexander Embiricos initiated an hourly poll on his X feed to assess how well Codex, an AI model, handles deep research-style queries. The poll reflects user curiosity about Codex's effectiveness in complex, investigative prompting scenarios. This setup enables real-time community feedback on AI research capabilities.

User Requests Screenshot of Error Messages in Embiricos' Hourly X Feed Poll

A user commented on Alexander Embiricos' X feed hourly poll, inquiring about specific error messages. They explicitly requested a screenshot for visual clarification. No further details on the errors or poll content are provided in the note.

Embiricos Actively Pursuing Bug Reproduction Steps from User Report

Alexander Embiricos responds to a user's bug report by confirming ongoing efforts to replicate the issue. He requests consistent reproduction steps from the reporter named Mitch Troy. This indicates active debugging engagement in a technical context, possibly software-related.

Embiricos Launches Community Poll on Codex AI's Communication Style Preferences

Alexander Embiricos is running an hourly poll via his X feed to gather user feedback on Codex AI's personality and communication style. The poll solicits preferences for conciseness vs. explainability and directness vs. collaboration, along with specific examples of effective or poor interactions. Responses are encouraged openly or via DMs, with a playful reference to providing "burnt toast" in replies for those familiar with the cultural nod.

ChatGPT-oWork-4.2-Codex-xHigh: High-Performance Variant in Embiricos' Hourly Feed Poll

Alexander Embiricos' X feed features an hourly poll highlighting "ChatGPT-oWork-4.2-Codex-xHigh". This appears to reference a customized or specialized version of ChatGPT, potentially optimized for work tasks with Codex integration and high performance tuning. The notation suggests iterative development in AI model variants tracked via social media polls.

OpenAI and NVIDIA Enable Enterprise-Wide Codex Rollouts with Proven Success

OpenAI partnered with NVIDIA to deploy Codex across entire companies, achieving successful results in initial enterprise pilots. Alexander Embiricos confirms positive experiences from deep integrations with multiple enterprises. The initiative invites other companies to participate in similar rollouts.

Alexander Embiricos X Feed Features Hourly Polls

Alexander Embiricos maintains an X feed that includes hourly polls. The content provided is labeled as a user note for this recurring poll format. This suggests a structured, time-based engagement mechanism on the platform.

Embiricos' Tool Hides Bash and Uses Automated Approvals Policy

The tool conceals bash commands and employs an approvals policy independent of manual bash approvals. Git functionality activates conditionally only within git projects. This design has persisted since at least 7 days ago.

Embiricos' Coding Environment Hides Bash and Uses Non-Manual Approvals

The coding setup conceals bash commands and employs an approvals policy independent of manual bash approvals. Git integration activates conditionally based on presence in a git project. This behavior has remained consistent since at least 7 days ago.

Enthusiastic Endorsement of Embiricos' Hourly X Feed Poll

A user poll on Alexander Embiricos' X feed, conducted hourly, receives a direct affirmative response of "yes!". This indicates positive reception to the proposed or ongoing poll feature. The brevity suggests strong, unqualified support from the respondent.

Windows Dominates All Non-Computer-Use Tasks in Embiricos' Workflow, Including Browsing

Alexander Embiricos reports that his entire workflow, excluding computer use tasks, runs exclusively on Windows—including browser usage. This preference highlights Windows as the platform for general computing activities outside specialized "computer use" functions. The statement appears in an hourly poll context from his X feed.

Codex Auto-Review Mode Streamlines Long Tasks with Reduced Approvals and Contextual Safety Checks

OpenAI's Codex introduces auto-review mode, enabling extended operation on tests, builds, and automations with fewer human approvals. A separate agent performs contextual checks on higher-risk steps prior to execution, balancing productivity and safety. Users like Alexander Embiricos now default to this mode for efficiency.

Embirico's X Feed Hourly Polls Slightly Accelerate Rate Limit Consumption

Alexander Embirico acknowledges that his X feed's hourly poll feature consumes rate limits at a faster pace than standard usage. More precise data on the consumption rate will be provided soon. This implies a minor performance trade-off for real-time polling functionality.

Embiricos Advises Temporarily Disabling Approval Mode in Config for Imminent Fix

Alexander Embiricos announced a temporary workaround for an unspecified issue in his X feed's hourly poll feature: users should remove approval mode from their config.toml file. A permanent fix is scheduled for tomorrow. This directive targets technical users managing configurations, likely in an AI or automation context.