isn’t this a… software factory?
Laptops weren’t built for parallel agents.

Introducing @boxesdotdev (YC F26).

Run Claude Code and Codex in the cloud, on unlimited copies of your dev environment, with your existing subscriptions.

3 free months of Pro: boxes.dev/launch-yc
0
0
0
74
finally… my ios keyboard doesn’t try to autocorrect “subagents”!
0
0
0
61
flying to Thailand in two weeks with my wife, 4 year old and 10 month old 🇹🇭

any tips?
2
0
1
236
switching costs are the lowest they’ve been

churn has got to be on the rise

what are you switching from and to?
1
0
0
70
Replying to @blovedev
I switched:

netlify to vercel
typescript to rust*
0
0
0
34
what happens if you close the browser while an agent is doing something?

been working on that in Hashbrown

I’d like to pick up where I left off without accidentally doing the same thing twice

…still unreleased, but I’m excited about this one
0
0
1
62
👇

this is how I feel about chat ui
imo a lot of software is meant to be invisible.

the irony of building them is that you have to stare at it every day. easy to forget that for many of your users, the ideal experience is barely having to think about it at all.
0
0
3
126
we are working to confirm this leaked screenshot

you would be really stupid to build you own claude/codex/conductor/t3...

there is no value in this, no money to be made, and T3 code won. only a fool would even dare to start building something like this in 2026.
0
0
1
98
this is big
json-render now works in Chat SDK agents

generative ui for every chat platform

→ your agent picks the components from your catalog
→ slack gets block kit, teams gets adaptive cards
→ every spec is validated before it posts
0
0
2
104
agent ui… seems easy enough to build in a few days, right?

I don’t think so

it feels like auth - don’t roll your own or at least do so knowing this is a deep and complex system that requires you make it look and work seamlessly and easily
0
0
0
69
ok, not nice
Running a business is just realizing how dumb you were 6 months ago every 6 months
0
0
0
101
this is good so far.

but you really need daybreak access
Trending repository of the day 📈

security-audit-skill

A coding-agent skill for multi-phase security audits with independently verified, machine-readable findings

Last 24h: 2,088 ⭐
Total: 64,786 ⭐️
github.com/cloudflare/sec…
0
0
0
68
you know when you open the keyboard on your phone and the thing you were reading disappears?

been fixing that in threadplane this week

small thing, but for me this stuff matters as much as the fancy agent features
0
0
0
54
q: how do you get around admin and org policies to use these ai tools?

a: quit
0
0
0
77
Yes. And they billed annual without a whiff of a notice. Waste of money.

I’m sure there are clones, slops, and personal solutions. Small shops need something free and agentic.
why does Docusign still exist and why is it so absurdly expensive

horrible product, horrible UX, horrible pricing, all around really surprising that this product has been so sticky

please, anything else
0
0
0
75
Does local inference make an app offline?
Your model can run on-device while tools, image resolvers or telemetry still use the network.
Check those paths separately, especially before adding a hosted fallback.
0
0
1
75
does this show early signs that people will want to have their own buildable-customizable-generatable UI?

... can I just recall to mind home much I loved winamp...
all the posts about people hacking their TVs have made me so happy, I can't wait to mess with mine later
0
0
0
64
I would give a local model one small job before exposing a whole UI kit.
Our Chrome tests passed the single-component subset and still failed nested cases.
A larger catalog needs its own tests.
0
0
0
100
what are you building this weekend?
8
0
0
227
A wrong stop-token ID can leave you with half a JSON string.
WebLLM's Qwen3.5 fix is a good example.
Record the model config alongside the runtime version, and rerun your exact workflow after an upgrade.
0
0
0
70