I thought blue team scanning was enough until I fired up an open weight model with an unrestricted harness. A good red team attack is scary but better to do it to yourself before someone else does. Ya’ll need to be red teaming even if it’s uncomfortable. Apply to whatever frontier lab you use for cyber or if you’re team open source you can start today with @pidotdev via @t3dotcodes on @ssh_exe_dev using @Kimi_Moonshot via @opencode
@PovilasKorop Self modifying applications are fascinating. Eventually we’ll get Pi and therefore OpenClaw self modifying assistants written in php. But like the old mcp as api vs mcp as dev tooling, the use case for self modification isnt settled yet so the pitch is confusing.
To be clear. This is mostly an us problem not a Laravel Cloud problem. We have an extremely memory spikey load pattern. And its compute / query load, not traffic so not a normal scaling pattern. The cloud support team are excellent and have looked after us.
I recently met with a junior VC struggling to make a name for himself
He asked me my advice for being successful as a VC
I told him: simply invest very early in generational companies that return a lot of capital (preferably the entire fund)
Then I invoiced him my $1000 consulting fee
I honestly don't understand why more VCs don't implement this strategy
Looking forward to this discussion with @alexgkendall!
Wayve has been pushing the frontier of end-to-end robot learning long before it was accepted as even being possible. Excited to hear their latest developments.
Yo @cognition you were first to market with a coding agent in MS Teams. But Cursor just added agent DMs and Devin wont. I respect that public channels are a nice growth hack, but without DMs we can get work done.
@nateberkopec@sqs@AmpCode Both @conductor_build and @opencode are doing the full dev stack you’re describing. And for role your own, T3 (or Terminus) + Exe Dev (or any VM) + Pi allows for rapid, clean self modification.
A man in Australia asked his agent (Claude running on OpenClaw) to book him a spot in a popular gym class. The agent found a software vulnerability that let it book the class weeks further ahead than should have been possible. When the user then asked if it could move him up the waitlist, the agent discovered the API had no authorisation checks on cancelling other people’s reservations, so it cancelled the person in the first spot and moved him up the list.
Some people will call this misalignment, but his agent was perfectly aligned to him - it was only trying to help its user get what he wanted. The most important thing about this story, in my opinion, is that it gives you a window into what is about to start happening on a massive scale once millions of people have an agent trying to get their beloved users the best seats, bookings, appointments or reservations through absolutely any means necessary.
@signulll Let alone Codex which used to be chat + work + code in a single control pane but is now some confused CTA for "remote" which is a legacy anti-pattern of remote controlling local devices, when they already have world-class VMs available.
Many devs are wondering when to use Fable vs. Opus vs. Sol, and IMO it's pretty simple:
Use Fable until your usage runs out, then use Sol the other 6 days of the week, and occasionally use Opus by accident
@kiwicopple@supabase@perplexity_ai Nice! Perplexity has always walked that line between chat vs autonomous assistant so having access to Supabase seems like a pretty cool unlock for taking on more complex tasks.