powerful, affordable cloud VM fleets for agents 👉 https://xitter.cf/t.co/kx9WEgqzaY 👈 backed by @ycombinator (F26) & @fdotinc
agentic ocean
Joined October 2024
- Tweets591
- Following14
- Followers3K
- Likes1K
boat retweeted
how I defend your boat VMs from pirates
Live fire training is critical to keeping the ship safe out at sea and in port.
George Washington is deployed to the @US5thFleet area of operation in support of the @CENTCOM mission.
boat retweeted
this is how you should be stress testing your product before a launch
you spin up multiple bots with their own unique personalities on @mausbot_
give them their own @boatdotdev computers
then let them loose on your product, all at once, each attacking it from a different angle
and you watch every one of them on a live canvas
find what breaks before your customers do
boat retweeted
💸 today I invested in a new thing:
cpu compute is essential to all personal agent services. every muse, grok bot, & instinct session uses a cloud vm.
however, it's scarce. hyperscalers didn't plan their datacenter buildouts expecting a world where every earthly human is using dozens of cpu vms for ai agents.
my friend @AniC_dev founded @boatdotdev to solve this. he's locking in bulk cpu compute prices now so he can provide the best full vm sandboxes for years to come.
before this he was working on arbitraging hetzner (what was previously the cheapest compute on the market), visiting sf a ton, & building an audience of developers here on X. once i realized that cpu compute is going to become as scarce as gpu compute is already (after joining morph) i asked to invest immediately.
there's no longer a moat in software or intelligence, the only durable thing is compute ownership. if you agree with this then you should find ways to invest in things like boat such that you're long compute.
follow @boatdotdev to see what they will make
weekly billable usage growth on boat.dev cloud VMs is crazy
agents spawn more and more VMs but also make them run for longer than before
that billable usage being so much cheaper than on Modal, Vercel or Fly means we're saving our users 6 figures every week
boat retweeted
Saved 3+ hours on a large build by using 15 @boatdotdev VM machines in parallel, for less than a dollar... massive productivity unlock. Building out a skill to just use boat images for all CI/Builds in dev.
CI/Build bottleneck solution activated.
@AniC_dev
boat retweeted
beyond stunts, daily usage of boat.dev VMs for serious agentic usecases is growing like crazy
we're not only the best cloud VM infra for agents we're also the most affordable at scale
if you need cloud VMs for agent fleets there is literally nowhere better to go
spinning up 700 boat.dev cloud VMs and gpt 6 agents that play Minecraft multiplayer on them
700 desktops + codex + game instances boot & start playing from a disk snapshot in <10s on 2.8k vCPU 5.6TB of RAM
total VM cost for 15 minutes of this: $6.45
boat retweeted
Say Good Bye to GitHub-hosted runners 👋
We moved our CI to @boatdotdev sandboxes to save 75% vs. Github runners AND cut out deployment time by 50%:
⚡ Unit tests: 7.1 → 4.0 min
🧹 Lint: 8.7 → 4.7 min
✅ 8,499 tests, now moving at lightning speeds
It took just a few hours with Opus 5.5.
Our GitHub Actions bill was on pace for ~$430/month.
That's over $5,000/year, and every backend PR still took 9 minutes to go live.
Now each CI job gets its own fresh sandbox, at 70% less cost.
It's unbelievable how fast we can experiment and swap code out!!!!
boat retweeted
The most underrated marketing move right now is optimizing for the AI agent holy sh*t moment
Nobody reads your docs anymore, they let their agents do that for them. And when an agent compares you against 5 competitors, it should see you as the best choice for a single reason.
@boatdotdev is one of the best examples I've seen. The whole moat they have, is that they're affordable, and they made sure every agent finds this info super easily:
- They make it super easy to see the pricing (aka their moat). 4 vCPU / 8 GB VM for $0.036/hour, billed per second
- They also have compare page prices the same machine on 15 other providers. E2B and Daytona are 9x, Modal is 21x. This lets the agent compare you against competitors even easier than before.
- All of it is plain text with an llms.txt, so an agent can read it
When I was personally comparing sandbox options for my app, I just threw in a link of boat, and my agent said "and it's billed by the second. That's kind of insane.". That's the moment you want agents to have on your product.
You can feed this post to your agent, and have it find one thing about your product that you will push heavily. This needs to be the reason why people choose your product over everyone else.
boat retweeted
been using codex + @boatdotdev to run and take screenshots of all my old projects. hundreds of screenshots so far for just 6¢ in sandbox usage 😌
boat retweeted
we're saving our customers >$7.5M this year
boat.dev provides cloud VMs fleets to agents at 9x cheaper rates than E2B & Daytona!
rn we're serving more than 12k vCPUs and 24TB of RAM to our customers, growing 7.5x in just 2 months
at $0.009/h per 1vCPU x 2GB multiples,
that's >$7.5M saved this year for our customers
imagine paying millions too much just to be hacking around using incumbents that don't even optimize for long running or always on agents fleets
boat retweeted
we casually doubled in just 7d how many full VMs are spawned daily on boat.dev
but the infra is holding stronger than ever!
status.ascii.dev
boat retweeted
Meet Otto 👾
Close the laptop. Your coding agents keep going.
Otto runs them on a cloud sandbox you own, and you steer them from your phone.
> Built with Tauri 2: Mac, Windows, Linux, Android
> Bring your own model keys
> Bring your own sandbox provider keys
> Permission prompts as phone notifications
> Free, MIT, no account, no backend
Works with @e2b @daytonaio @boatdotdev @boxd_sh or your own machine
Launching soon 👀
boat retweeted
introducing LLM & API credentials proxying in boat.dev
when ON, sandboxes get mock API keys & subs that route to yours safely
it allows to:
-> let ur subusers or team use AI in credentials-less VMs
-> monitor their token usage & stop them any time
you can also now:
- rotate keys / subs at any time (for ex, if you run out of usage)
- setup multiple AI credentials profiles for tidyness
- create team environments and secrets
boat.dev, the cheapest, most powerful and practical sandboxes for you, your agents, your teams, and your users
boat retweeted
If you pay for UGC, it isn't UGC, it's just a commercial shoot.
UGC is organic, it's earned.
The fellas at @boatdotdev right now are killing it with their organic UGC.
Real UGC is sticky, it's convincing, it's unique, it's often low effort too.
A user takes a screenshot and says "wow, this is awesome" is UGC.
A user making a video walking through how to set up their sandboxes and connect their agents is UGC.
Someone posting about how much time they've saved because of a product and how thankful they are they found your company is UGC.
It's supposed to be raw and organic.
Not polished, scripted, and paid for.
Paying a creator to use your product and make a video and calling it UGC is disingenuous at best, and illegal at worst.
The only reason homie was able to get away with this is because it was an ad program, because the consumer knows it's an ad.
boat retweeted
we made a RTS game for our YC company boat (F26)
⛵ control an armada of boats (the best AI sandboxes in the world)
🔥 destroy the Modal castle protected by Daytona walls and E2B Towers
let's kill it in the sandbox market
boat.dev/game
boat retweeted
Worked on getting sandboxes into T3 Code
Instead of running on main or a new worktree, it spins up a fresh boat.dev VM, run the agent there, and adds it as a remote environment on all devices
No more managing memory on my machines or balancing worktrees
