Plugins, MCP, Skills at @OpenAI.
Prev Director of Engineering @Shopify, co-founder and CEO @Stellate (acq. by Shopify), co-creator of styled-components.
one week later: good news and bad news on chatgpt plugin submissions.
👍 good news: we've nearly doubled how many reviews we completed this week by tweaking our reviewer setup, so many more of you have heard back from us.
👎 bad news: our backlog is still growing because submissions have kept growing exponentially.
i said we'd catch up to the backlog in a week. we haven't, and i'm sorry to all of you who are still waiting.
we've now started fundamentally rethinking how we review plugins so we can deal with this new plethora of plugins, but it'll take some time to get it done.
thanks for bearing with us 💖 i will keep updating you all as the situation evolves.
let's talk about submissions. since the beginning of the year, chatgpt plugin submissions have increased by almost 10x.
then, since the explosion of interest in plugins after devday, they 3x'ed again! ��
it's been really difficult for our review team to keep up, and it shows in the review times. right now, the eta for you to hear back from us is about 2 weeks – that's too long!
so, we've declared a temporary code red. we're relaxing some of our nice-to-have requirements so that we can increase throughput.
it's going to take us about one week to catch up with our backlog, so if you have submitted so far you'll hear back in the next days.
we're also:
1. going to show the eta in the submission portal so at least you know when you'll hear back from us
2. working on automating more review checks so you get immediate feedback without having to wait for our review team
3. taking another look at the requirements to make sure we check for only the most critical things to keep our users safe
keep building plugins and keep submitting them; we're hard at work making that a great experiencer. thanks for bearing with the team.
more context in the video:
MCP Apps ❤️ WebMCP
Great job by @ndmccormack , @mxstbr and the ChatGPT team! This is the future of interfaces, the chat can now interact directly with MCP Apps, using WebMCP.
Making the agentic web seamless ✨
I'm excited to share that WebMCP is now enabled for MCP Apps in ChatGPT Desktop! 🚀
@OpenAI is also collaborating with https://t.co/WAjlykp7dX to make WebMCP the default way for agents to interact with MCP Apps.
Another big step toward a more seamless agentic web!
"Almost exactly a month ago, Foursquare was acquired, marking the end of a 17-year story that started around my kitchen table in the East Village in 2009.
The Foursquare acquisition was a $0 outcome for equity holders – $0 for preferred, $0 for common. It’s bittersweet to see these details surface publicly – at least now there’s an opportunity to comment on them."
Man.
https://t.co/dvxjuSL39T
I’ve been getting a ton of DMs and emails about plugin submissions. Sorry if I haven’t replied to yours yet.
We love seeing all the ways you’re making ChatGPT & Codex more useful!
We nearly doubled reviews last week, but submissions are outpacing us. Thanks for your patience 🙏
@frantzfries agreed. we actually split them already & code updates go live automatically (every day roughly). need to do more automation for metadata changes!
let's talk about submissions. since the beginning of the year, chatgpt plugin submissions have increased by almost 10x.
then, since the explosion of interest in plugins after devday, they 3x'ed again! 🤯
it's been really difficult for our review team to keep up, and it shows in the review times. right now, the eta for you to hear back from us is about 2 weeks – that's too long!
so, we've declared a temporary code red. we're relaxing some of our nice-to-have requirements so that we can increase throughput.
it's going to take us about one week to catch up with our backlog, so if you have submitted so far you'll hear back in the next days.
we're also:
1. going to show the eta in the submission portal so at least you know when you'll hear back from us
2. working on automating more review checks so you get immediate feedback without having to wait for our review team
3. taking another look at the requirements to make sure we check for only the most critical things to keep our users safe
keep building plugins and keep submitting them; we're hard at work making that a great experiencer. thanks for bearing with the team.
more context in the video:
the chatgpt plugins team is coming to MCPcon in san jose this week! ✋
we'll be giving a few talks and you'll see many of us roam around the conference. come hang out, talk MCP, chatgpt plugins, submissions (🙈), and show us what you’re building.
we’re also hosting a happy hour on oct 22, 8–10pm. even if you're not going to the conf, come hang out with us:
https://t.co/hxVzk23doX
@tweetsbycolin@_Dapp@devingould18 it’ll restart right now :| we have to fix that.
But good news: updates can go live automatically, so once you’re approved you can push an update to your MCP and it should go live automatically within a day!
Yes it's true. We use code mode & tool search across ChatGPT and Codex now.
Our current limit of tools per MCP server is ~2k, but that's really only there to prevent abuse. We have plenty of MCP servers that are in the hundreds and even some in the thousands. Even Cloudflare is about to serve their raw toolset of 3k tools to ChatGPT instead of code mode!
Go watch @sarfata's keynote at MCPCon this week for more fun stats :)
there's a nascent trend of hiding MCP tools behind two generic tools: search & execute.
i don't think it's a good default. this basically means implementing tool search & code mode on each MCP server.
but clients like codex already have tool search & code mode that a/ work across all MCP servers and b/ we train our models to be great at!
why make every MCP server build that layer and then make the model navigate both, including a custom one on each MCP server that it isn't trained to be great at?!
expose your actual tools and let clients* find and compose them.
*clients that support tool search & code mode like ChatGPT and Codex