GrokBot? You mean AuthSimulator2026?
As they say in the medical field:
Delulu is the solulu
Repetitive negative thinking is linked to Alzheimer’s protein buildup in the brain.
Them: Grinding a 6th day to get 10-20% extra juice against 100 identical products
Us: Relaxing, vibing, journaling to unearth unique, 100x outcomes buried deep beneath the surface
Oh, young, misguided founder, this is not the way
Done right for an organization, whenever you 1.5-2x headcount, you will 10x throughput
That is the goal
It means you are hiring great people, the systems are building leverage, and you are building the right things
Show more
Reliable way to 10x your output:
- Say no to 50% more things
- Figure out your own internal bottleneck
- Hire for what you can delegate
Repeat. Infinitely
Ironically, we’ve been doing “Sandboxes” since 2020. Fast, Dynamic, persistent compute
Now combine this with cloud agents, live branches (🔜), feature flags and observability
Unleash every agent, in parallel with zero friction, to build anything instantly
Show more
Announcing: Railway Sandboxes
Spin up thousands of VMs right next to your infra on Railway
Run, checkpoint, and fork and let your agents (and code) run
Reliable way to 10x your output:
- Say no to 50% more things
- Figure out your own internal bottleneck
- Hire for what you can delegate
Repeat. Infinitely
Plant the tree today. Take life seriously today. Do the thing today.
Your future self will thank you (instead of hating you)
there really has never been a better time to take things seriously
Some people make their own luck
Because luck, in some part, is the ability to steer outcome via precision
Close enough
Welcome back "Community Adjusted EBITDA"
JUST IN: Anthropic says they’re highly profitable if you take out some of their biggest expenses.
@ivanburazin Tech consistently forgets that most things are not technology problems, they are human psychology, habit, and risk tolerance problems. They take time.
This makes tech often right about the thesis, and wrong about the timeframe over which it’s realized.
Show more
Earlier this year, this would have probably paged us
But we only found out about this because...someone was randomly looking at DataDog
~5m RPS on the Railway network right now
The infra team has been putting in insane work to scale at the pace needed to meet demand from AI
All systems go 🟩🚅🟩. Onward and Upwards
Show more
My biggest life pro tip is buy 5x the amount of consumables you use and slap em in various places
Toothbrush, charging cable, deodorant, etc. In your car, in your luggage, in your day bag.
Then you never think about it again; it’s awesome
Show more
"Engineers who used X also used Y"
Reaching levels of "Landing Page Logo Testimonial" scientists previously thought impossible 😂
Boutta crank bass to 10, queue up Flo Rida, and godspeed to whoever takes that chariot next
You can adjust Waymo EQ?!
Many folks mistake pedigree for excellence
It’s a proxy but it’s not the whole picture
Miss that, and it’s a costly mistake
Railway now has Cloud Agents
Fan them off, from MCP/Mobile/etc, and run them alongside production infrastructure
Next up: snapshot anything, make a change, and have them walk it into production
You can just build things, remotely
Cloud Agents are now in Priority Boarding: real VMs running on Railway Metal
Fire off sessions in harnesses like Grok Build, Claude Code, Codex, and more
Close the laptop and pick sessions up from anywhere: Dashboard, CLI, or Railway mobile
Show more
Why don’t they just make the whole S3 out of NVMe?
So caching everything on NVMe is good for read throughput I guess.
You can disagree with any single take. But the pattern here matters:
@tobi wants Canada to win. He puts his ideas out in the open, takes the heat, and keeps building one of the most important companies in the world from 🇨🇦
That is what ambition for your country looks like.
Show more
You have to open yourself to the possibility of mistakes, failure, etc
The moment you stop being open to mistakes, life closes off
If you open yourself to it, life opens up. It’s that simple