just got access to abliterated chinese models on @morphllm for cyberdefense
prompted to pentest us for:
> cross tenant secret leaking, penetration or spying
> guest-host takeover
> control pane takeover
ngl our engineering is good, it's struggling, we're gonna need real pros
prompted to pentest us for:
> cross tenant secret leaking, penetration or spying
> guest-host takeover
> control pane takeover
ngl our engineering is good, it's struggling, we're gonna need real pros
24
2
149
7,548
i don't think you understand how crazy this is
you can save literally millions per year and cut costs by 80%
while getting work done on the best CPUs and VM platform on earth
you can save literally millions per year and cut costs by 80%
while getting work done on the best CPUs and VM platform on earth
on boat.dev cloud VMs on a $250 budget you can spin up in <2s from a snapshot:
> 17k typical CI runs (6.5x cheaper than on blacksmith)
> 31k RL runs (8x cheaper than on daytona)
> 868 coding agents for 8hr
on 4c x 8gb x 50gb nvme, ryzen 9 full VMs
5
2
54
3,030
gm sf
came to the attention of @utpalnadiger that i had not worn a tie in 2 days
which is utterly undignified behavior from a YC founder and captain of boat.dev
so today we fix that
came to the attention of @utpalnadiger that i had not worn a tie in 2 days
which is utterly undignified behavior from a YC founder and captain of boat.dev
so today we fix that
8
1
44
1,992
counter-intuitive fact about providing cloud VMs / AI sandboxes:
letting users spawn lots of short lived VMs costs way way way more than letting them run a few big VMs for long
time is cheap, ellasticity is costly
and we're the only ones not lying about it!
benefit of our explicit limits and pricing is that long running VMs cost way way less on us and burstsy short lived swarms remain competitively priced
letting users spawn lots of short lived VMs costs way way way more than letting them run a few big VMs for long
time is cheap, ellasticity is costly
and we're the only ones not lying about it!
benefit of our explicit limits and pricing is that long running VMs cost way way less on us and burstsy short lived swarms remain competitively priced
boat offers the least expensive full VMs for agents at $0.00001/s for 4c 8gb ram 50gb nvme
to further protect these prices historically we've had to cap the n° VMs starts / day
we raised the limits 2 times already, and today we add a way to burst above them, up to 3x, using get up to 25% free credits for starts, snapshots & VM time by buying more of them upfront
the kind of spiky workloads we saw in prod that benefit from bursting remain 4x to 8x less costly on boat.dev than Daytona, E2B or Modal
and it's now way less frustrating to pull them off
on boat.dev more than anywhere else, the generous compute pricing aligns closely & sustainably with real costs
learn more:
docs.boat.dev/limits#limits-…
8
1
46
3,174
boat offers the least expensive full VMs for agents at $0.00001/s for 4c 8gb ram 50gb nvme
to further protect these prices historically we've had to cap the n° VMs starts / day
we raised the limits 2 times already, and today we add a way to burst above them, up to 3x, using get up to 25% free credits for starts, snapshots & VM time by buying more of them upfront
the kind of spiky workloads we saw in prod that benefit from bursting remain 4x to 8x less costly on boat.dev than Daytona, E2B or Modal
and it's now way less frustrating to pull them off
on boat.dev more than anywhere else, the generous compute pricing aligns closely & sustainably with real costs
learn more:
docs.boat.dev/limits#limits-…
to further protect these prices historically we've had to cap the n° VMs starts / day
we raised the limits 2 times already, and today we add a way to burst above them, up to 3x, using get up to 25% free credits for starts, snapshots & VM time by buying more of them upfront
the kind of spiky workloads we saw in prod that benefit from bursting remain 4x to 8x less costly on boat.dev than Daytona, E2B or Modal
and it's now way less frustrating to pull them off
on boat.dev more than anywhere else, the generous compute pricing aligns closely & sustainably with real costs
learn more:
docs.boat.dev/limits#limits-…
15
4
91
7,654
I'm so sick of people submitting shit/fake security issues found by subpar models with total slop reports just to beg for jobs, credits, bounties or attention
above being not authorized, not tested, often not impactful/exploitable in practice or pure imagination sometimes
they're made to look scary at first and lose my precious time trying to secure our infra as much as can be AGAINST REAL POTENTIAL VECTORS!
above being not authorized, not tested, often not impactful/exploitable in practice or pure imagination sometimes
they're made to look scary at first and lose my precious time trying to secure our infra as much as can be AGAINST REAL POTENTIAL VECTORS!
22
2
93
4,443
Anicet retweeted
Created a Triage agent this week inside Squad.so...
I get feedback from so many places like Slack, emails, customer calls, X DMs, WhatsApp messages, Production Logs, Session Recordings, Analytics etc.
I am getting lost navigating between all of this feedback and ne, what is not done, which takes priority etc.
So I created a new agent inside Squad - for triaging all this feedback and combining everything into a dashboard that I can come back to everyday.
This is the end result. It collects every single feedback item for all the connected sources, deduplicates it and turns it into feedback threads and even assigns a priority to it.
My Squad agent created this and only I can see it from my Tailscale connected device.
The next step is for agents to read this feedback and automatically open PRs for each of these based on assigned priority and approval.
Going to eventually turn this entire process into a closed loop so that it can run on its own as much as possible (with me in the loop minimally)
I get feedback from so many places like Slack, emails, customer calls, X DMs, WhatsApp messages, Production Logs, Session Recordings, Analytics etc.
I am getting lost navigating between all of this feedback and ne, what is not done, which takes priority etc.
So I created a new agent inside Squad - for triaging all this feedback and combining everything into a dashboard that I can come back to everyday.
This is the end result. It collects every single feedback item for all the connected sources, deduplicates it and turns it into feedback threads and even assigns a priority to it.
My Squad agent created this and only I can see it from my Tailscale connected device.
The next step is for agents to read this feedback and automatically open PRs for each of these based on assigned priority and approval.
Going to eventually turn this entire process into a closed loop so that it can run on its own as much as possible (with me in the loop minimally)
15
4
28
3,298
