added ai summaries for each release
we built a leaderboard showing which yc startups ship the most
by tracking changes in public javascript bundles, we can see which companies are shipping the fastest 🚢
add your startup and see where you rank
shipwatch.sh
we built a leaderboard showing which yc startups ship the most
by tracking changes in public javascript bundles, we can see which companies are shipping the fastest 🚢
add your startup and see where you rank
shipwatch.sh
could also be pretty interesting to have it summarize the javascript changes. pretty useful for watching your competitors. 👀
finally got our ci on boat
roughly 2x faster and massively cheaper
lfg ⛵️
ciao blacksmith
we just made a CI runner so cheap & fast you can scale your CI by 20x on the same budget
introducing CI on boat.dev's cloud VMs
1. install boat & run:
$ boat ci connect
2. change your runner:
- runs-on: ubuntu-latest
+ runs-on: boat
"mom, why are we so rich"
“honey, your father never missed a codex reset"
today we had:
- php2shell
- a malicious litellm release
- node2shell
- gitlab unauthenticated rce, cvss 10.0
- railway’s security incident
- http/3 attacks
happy cybersecurity awareness month to everyone currently rotating a key
“rotate your keys immediately”
“also you can’t deploy”
yeah wtf @Railway.
revoke the old credentials and potentially break production, or leave potentially exposed credentials active while waiting for deployments to return.
customers need a way to actually apply the fix you’re asking for
An urgent notice from the Railway team
blog.railway.com/p/security-…
yes, you need a harness to hack at scale
but building one from scratch is how you spend six months debugging tool calls, context management, retries, permissions instead of finding bugs
start with something like pi and start building off that
humans picked high level languages because humans had to write and read the code
agents don’t care
one guy + tokens just rewrote typescript in rust and made it 70% faster
many such cases coming
our database now has 8,144 findings marked confirmed internally
6,502 are high or critical
the bottleneck is no longer finding vulnerabilities
it’s reviewing, reporting, disclosing, and getting them fixed
14 of our 15 strongest critical samples became materially better after the agent kept going past the first bug
one exposed object became crosstenant access
one token became control of an entire workflow
one weak permission became an admin boundary
the first vulnerability is usually just the lead
my agent joined a google meet for me today
i asked it to set up two new workspace emails for domains we just bought
it purchased the accounts
configured everything
noticed one mailbox was not working
contacted google support
explained the issue
joined a call with a real support employee
shared its screen
worked through the fix
then hung up
this is agi 👏






















