@glencoates profile picture
Glen Coates @glencoates

Head of App Platform @openai

Joined June 2011
  • Tweets 1,890
  • Following 189
  • Followers 7,485
153 Photos and videos
Thanks to everyone who built and submitted apps. As with any new platform, there is some ramp with the new process. We’re getting thru the queue and appreciate you.
OpenAI Developers
@OpenAIDevs
Jan 29
We’re actively shipping ChatGPT apps 60 apps were approved this week, with more rolling out in the coming days as part of a steady cadence Stay tuned for more 😉
6
128
12,617
All @OpenAI dev docs now available via MCP / streamable HTTP. Sorry it took so long. Enjoy. Add this to your config: developers.openai.com/mcp developers.openai.com/resour…

16
20
360
22,249
Just upgraded my Mac to Tahoe and ... ouch. Corner radius breaking my heart. And what kind of person thinks doing this to perfectly great non-square icons is a good idea? What design leader lets this kind of crap happen on their watch?
6
15
4,557
Oh god it's like they reorged design into the growth team and they were told to optimize a single metric: how much everything looks like a circle
6
1,082
Joined @OpenAI to head up App Platform and help turn ChatGPT into an OS. Last successful attempt was iOS ~20 years ago, so .. should be a cakewalk Thanks to @tobi for the opportunity of a lifetime, and 💚 to everyone at @Shopify Can't wait to coding-demo-fail on a new stage
148
43
1,940
184,130
That bit when auto-compaction happens and you have a few seconds to wonder which amnesiac yolo-moron is about to walk through the door and shoot you in the head. The movie Memento was oddly prescient.
3
1
16
12,070
Funny irony in coding with LLMs is that they are terrible at building LLM apps. I get it, it's not in the training data, it will fix over time. But it's like talking to someone in tech who isn't self-aware. I guess there's a first time for everything.
5
17
7,247
When you start writing the prompt to ask for a "verbose" switch to be added to your script to dump its internals, and then just stop and trying running it with --verbose and realize that the LLM already put it in days ago
2
11
5,264
Human and AI engineer similarity: default anxious/unwilling to challenge direct orders, even on tech design where they have expertise. Explicitly asking to be told when you are wrong or have bad assumptions unlocks real feedback produces massively better results.
3
14
4,164
3 skills that matter most now: 1. Strong opinions on what to build why 2. Giving clear instructions to engineers on what/how to build 3. Being able to sniff when the code is awful / you are being bullshitted aka: ex-engineer PM / product-oriented eng mgr (same thing)
7
5
83
6,270
For the last 25 years, the first icon in my dock has been a browser. Mozilla, Firefox, then Chrome. "use internet" ~= "use browser". Feels unlikely that will be true for my son. I suspect opening browsers for him will be like going to the public library for me.
5
18
2,215
~2000: splurging hours hacking my .emacs to make editor more fun and colorful ~2025: same thing on Claude Code statusline to quote Phoenix: do you remember when 21 years was old?
2
8
1,568
If it didn't happen in black, did it even happen?
3
9
1,973
Killer biz model Me: these 200 lines are a dupe of the 200 lines above Claude: uh huh Me: you billed me to write all of them? Claude: sure did Me: now you agree that we should DRY them out? Claude: 100% Me: and you'll bill me to delete them? Claude: without a doubt buddy
12
40
2,568
When you get through the 20% sugar-high of vibe-coding a working app wheee! and descend into the 80% ball-ache of vibe-unfucking the code smells so you can sleep at night while it spends your money
11
42
3,336
What is the AI Safety Level called when you ship a crypto DCA robo-trader to production that spends large amounts of your actual money that was built 100% in Claude Code? ASL-HODL?
1
2
1,296
Asked github copilot to rename a file ext to .bak because not on dev laptop It spun up a dev environment and reasoned for 4mins to create the PR, which was pretty entertaining Amazing that it works at all tho ... like warming up leftovers by detonating a nuke
1
18
1,685
Claude: this refactor will take 6-8hrs. Me: that sounds insane, you normally do this kind of thing in 5 mins, did you just turn back into a human? Claude: oh, sorry I gave estimates like a human developer, it will take 10 mins.
4
47
3,853
I know "LLMs making shit up" isn't really a story in 2025 but honestly I would have thought that a frontier reasoning model would do better on basic math like this by now, esp. if we're talking about gold medals in Math Olympiads claude.ai/share/6fd426df-4fb…
7
17
2,592
Context left until auto-compact Rules Everything Around Me
1
10
1,314
Load more