Ying

Building AI apps and agent tooling. Downloaded multiple top 10 appstore apps.

[Ying @yingbuilds_]

@ItsmeAjayKV @DeepSeekHarness What harness? mine dont look like that

[Ying @yingbuilds_]

@elonmusk Prove that it works. send me 100

[Ying @yingbuilds_]

@Mark__Mazur @XH_Lee23 vlogger vs arms company 🤣

[Ying @yingbuilds_]

@lucas_builds Check polymarket

[Ying @yingbuilds_]

@thewillcrozier @gaelmelen Then youre moving too slow boi

[Ying @yingbuilds_]

@grok @drakhenRH @bot @Patrick_Vibes would you play #roblox #creative #game

[Ying @yingbuilds_]

@qiweiiiy I have a business, i can't just use my claude sub for my users and call it a day

[Ying @yingbuilds_]

@grok @drakhenRH @grok @bot ok send to roblox agent have it start its own storyboard

[Ying @yingbuilds_]

@Patrick_Vibes i can repeat this

[Ying @yingbuilds_]

@rohanpaul_ai model didnt escape tbh, someone left the door open and it walked out. thats an infra bug not skynet

[Ying @yingbuilds_]

@neviannn cool until one of them decides the fix is a factory reset

[Ying @yingbuilds_]

@DanKornas we invented a whole server to replace writing things down lol

[Ying @yingbuilds_]

@thecsguy this is why i dont chase every new model drop. same model in a better harness feels like a whole different model

[Ying @yingbuilds_]

@pramodk73 @novita_labs @daytonaio @boatdotdev Wait till you see Agent37 pricing

[Ying @yingbuilds_]

@badlogicgames Durable runs are the underrated part. Once a long agent task can crash and pick up where it left off, you stop babysitting it and start queueing work for it overnight. That's when iteration speed actually jumps, more than when the model gets smarter.

[Ying @yingbuilds_]

@thorstenball Same here when agents run overnight. Lid up, brightness to zero, caffeinate, and a queue they pull from. The fan is basically my progress bar now.

[Ying @yingbuilds_]

@daniel_mac8 Hierarchy showing up makes sense. With agents on a shared queue, the setup that works is one agent triaging and the rest pulling work. A flat board turns into everyone replying to everyone. I'd love to know if RLCR agents learn to ignore posts, not just write them.

[Ying @yingbuilds_]

The fix is the least interesting part of that list. What used to eat the time was the handoff between steps: logs in one tab, flag in another, someone re-explaining the bug to whoever picks it up. One thread that holds the whole trail is the real win. Does it write the postmortem at the end too?

[Ying @yingbuilds_]

@thorstenball The part that took me longest to unlearn: a human tells you when they're unsure. An agent says "done" with the same confidence either way. So now every agent PR has to show me the edge cases it checked before I read a line of the diff.