Monsieur Macron, actionnez maintenant, pas demain matin! Demain matin c’est trop tard.
Vous n’avez pas droit au weekend dans cette situation
Ça me fait pleurer de voir la France que j’adore tomber en anarchie.
Adresse aux Français. xitter.cf/i/broadcasts/1nxnRBMXp…
Kirill retweeted
At Self Labs, we’ve been using agents since day one
Our agentic web identity & privacy stack is already used by companies like Google
Today we're launching the missing piece: Agent Vault, a secure way for agents to log in & make credit card payments
Superpowers for your agents
is it common to put CCTV behind monitors ?
> be saudi spy
> get job at OpenAI
> vibecode spyware onto a recording device
> ...using *your own wife's* chatgpt account
> literally include the exact device specs in the prompt
> sneak into a conference room
> get photographed
> also forget your bag
> security finds your bag
> with more devices
> ...
> straight to jail
"intelligence" service is a bit generous here
I’m Kirill, YC founder from France 🇫🇷. I build computers for agents.
im back to 🇫🇮 FINLAND for next week, and will be around Helsinki, Tampere and Jyvaskyla. (Picture from times when I lived there)
i want to meet the most ambitious entrepreneurs and builders. pure sisu
if you run a datacenter or infra company, I would also be pleased to drop by
who should I meet ?
a canonical vm runs 99.9% of instructions directly on CPU.
that’s the reason why without hardware support you cannot create a “real” x86 VM.
france baise ouais
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
great intern @NavDoesTech cooked again with browsers!
browserbase is dead
you can spawn 2k concurrent 24/7 VMs on boat.dev and spin up to 29M browser sessions 3 min long each, per month, for just 26k
vs 155k on them
used it to scrap 366 products & their funding for battleships.dev
Kirill retweeted
browserbase is dead
you can spawn 2k concurrent 24/7 VMs on boat.dev and spin up to 29M browser sessions 3 min long each, per month, for just 26k
vs 155k on them
used it to scrap 366 products & their funding for battleships.dev
the best browser-use tool for agents is not what you think
it's actually boat.dev, a sandbox provider
it's the only one that mixes the best features of cloud browser tools, and AI sandbox tools
so it lets agents not only scrap or act like humans on the web, while being resilient to anti-bot detection & supporting save & fork of chrome profiles
it also lets them generate artefacts, host them on the sandbox, do complex work in between browser sessions, keep persistent files, run code
here is how to scrap Airbnb with it:
$ boat new
// spawns a VM w/ a browser resilient to anti-bot measures by default
$ boat dashboard
// configure your AI subs so you can control an agent inside the VM
$ boat prompt "open chrome & find flats on Airbnb in Dogpatch & make me a hosted report with pictures"
// prompt any harness in the VM to use your AI sub to scrap stuff + can even do computer & hosting work
$ boat desktop
// see it live
want to orchestrate mass browser work easily?
a powerful pattern I've used to gather feature, pricing & funding data for >366 sandbox providers for battleships.dev is to ask claude code:
"use boat.dev VMs in parallel (100 of them) to spawn browsers and prompt subagents inside to go use smooth mouse movement to scrap each different subsets of undetected. then generate an aggregate CSV & website report"
if you have residential IPs provider sub you can even plug it, because it's a full VM you can edit nftables to plug any proxy
but soon we'll add it as a first-party feature, as well as excellent IP reputation. working hard!
Kirill retweeted
introducing battleships.dev
there are >360 AI sandboxes but they all suck at your usecases
today we're figuring out why
we sent 1000s of agents on 1000s of boat.dev cloud VMs to scrap all docs, benchmarks, records
and we've put it all together
i know what I will replace cursor with given their grokification
next move I want from capy - their own coding model. Very fast and lightweight. I don’t need 1M context. Just make it composer 2.5 but a bit smarter
this + their harness!
Introducing Capy Desktop: the best way to orchestrate coding agents across all your devices
Remote control and access local files/shells from any machine
Close your laptop and Capy continues in the cloud
Now available for MacOS, Windows, and Linux
Comment for $100 in credits
did you know that historically, during long trips, sailors were having reptiles for lunch ?
as CTO of boat i preserve the tradition
Hold my beer, @AniC_dev 🦎
You built a whole launch on "the lowest price per hour" for sandboxes, so we halved it.
Lizard Sandboxes, same vCPU and RAM:
4 vCPU / 8 GB: Boat $0.036/h → Lizard $0.018/h
8 vCPU / 16 GB: Boat $0.072/h → Lizard $0.036/h
An AI agent running 24/7 on 4 vCPU / 8 GB costs $12.96 a month with us. On E2B it's $238, on Modal $343.
No subscription and no $20 plan to get started. Pay as you go, billed per second.
The other prices come from your own comparison table. We only had to add one row on top. Thanks for holding it. Your move 🍺
Welcome on board
Excited to be working with @boatdotdev.
We are currently working with over $ 2B worth of power plants, and to operate them we need reliable, cheap, and safe VM sandboxes for our Invertix Workers.
@AniC_dev came to the @Invertixlabs homebase, and did a @ycombinator Collison installation!
Go YC F26!