SpaceXAI engineer, ex-Cursor:
"my bots cleared $96k last month. my last job paid me $140k for the entire year
i open the system about 12 minutes a day, almost always just to say yes or no
one person, 19 agents, and a memory file worth more than the model they run on"
in a 29-minute session, a SpaceXAI engineer shows the exact roster and how the work moves between the bots
the model is rented. the memory is the only thing you actually own
watch it, then build that memory into your own team with the 12-step system in the article below
SpaceXAI engineer, ex-Cursor:
"my entire company costs $340 a month to run, and it has been profitable for 11 straight months
27 agents do the work, and every one of them reads the same memory file before it touches anything
i tried it without that file first and burned 5 months rebuilding the same context every morning"
in a 36-minute episode, a SpaceXAI engineer walks four levels of Grok Bot, from a single bot to a team that ships on its own
the agents are a weekend of work. the memory is why the company still exists in month eleven
watch it, then build the same memory into yours with the 12-step system in the article below
SpaceXAI engineer, ex-Cursor:
"i have not written a prompt since june, my agents write, test and fix their own now
18 of them run in parallel, they closed 1,700 tasks last month, and their error rate is down 74% since i stopped touching them
$2.4M in twelve months, headcount of one, and the only thing i ever scaled was a 400-line memory file"
in an 83-minute workshop, a SpaceXAI engineer builds a self-improving agent team from scratch
agents that get better on their own are not a model upgrade, they are a memory design
watch it, then build that memory into your own team with the 12-step system in the article below
Jimmy Ba (Co-Founder of xAI and GrokBot):
"9 out of 10 engineers at xAI now work through agent loops instead of writing the code themselves
that one shift is why they ship roughly ten times faster than a year ago
every engineer here keeps ten to twenty agents running just to handle their own routine"
in a 20-minute talk, an xAI co-founder lays out where AI engineering is heading next
the part he leaves out is what stops twenty agents from drifting apart by week three
watch it, then build that memory into your own fleet with the 12-step system in the article below
Ok he really just f*cking said it out loud on camera.
The CEO of Anthropic just said "for too long the industry lied" about how dangerous this technology really is.
He says the risks were dressed up as positively as possible for years, and that fixing any of it starts with telling the truth.
And what makes it sharper, he made a point of saying Anthropic never did.
All of this landing just days after he published an essay arguing that the entire industry needs to slow down.
He is pointing at everyone else.
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: https://t.co/OGyPb7yaYt
SpaceXAI engineer, ex-Cursor:
"i replaced a project manager, a designer and a developer with three bots and a group chat
we shipped 9 features last quarter, and i sat in zero meetings to do it
it's a $1M run rate with one name on the payroll, mine"
in a 29-minute session, a SpaceXAI engineer builds a full team of Grok Bots from scratch, with one agent managing the rest
the bots were never the hard part. keeping them from forgetting what you decided last week is
watch it, then build the one-person version with the 12-step system in the article below
SpaceXAI engineer, ex-Cursor:
"my bots fix 100% of my bugs now, i haven't opened a debugger in months
i keep 20 to 30 of them running at once, and the whole setup took one afternoon to wire
then people started paying me to build theirs. 11 of them at $4k each, in two months"
in a 27-minute walkthrough, a SpaceXAI engineer builds his real Grok Bot setup live, start to finish
everyone is racing to use this thing. almost nobody noticed you can get paid to set it up for them
watch it, then see which 5 builds people are already paying for in the article below
SpaceXAI engineer, ex-Cursor:
"i haven't hired a human in a year, i just spin up another bot
50 of them run on a $200 plan, and they clear what used to be a 3-week sprint before i finish breakfast
we did $1.2M this year on a headcount of one"
in a 54-minute workshop, a SpaceXAI engineer shows the setup that runs 50 bots off one plan
this is the piece almost nobody builds, and it's why their agents fall apart past three
watch it, then build the same memory into your team with the 12-step system in the article below
SpaceXAI engineer, Matt Palmer:
"i don't run 100 agents, i run the one that runs the other 100 for me
i haven't opened a task myself in weeks, they clear a 6-week backlog overnight and ping me maybe twice a day
we did what a 40-person team does last month, on a headcount of one and a $300 plan"
in a 37-minute workshop, the engineer behind GrokBot shows the orchestrator setup his own team runs on
this is the part almost nobody copies, and it's why their agents rot by week three
watch it, then build the same memory into your team with the 12-step system in the article below
Creator of GrokBot:
"the version of GrokBot i actually run isn't the one you're using
i've got 30 agents writing their own prompts now, i haven't typed one in months and they ping me maybe twice a day
we shipped 2,000 PRs last quarter with 6 people, my old company did half that with 60"
in an 84-minute walkthrough, the creator of GrokBot shows the exact setup his own team runs on
this is the part almost nobody copies, and it's why their agents rot by week three
watch it, then build the same memory into your team with the 12-step system in the article below
a 9B model learned to drive from scratch on 5,500 hours of data collected by regular people, and beat every open driving world model in Valeo's comparison.
that's VATIX 9B, trained by @Valeo_Group. all 5,500 hours of its initial training came from @natix_network, everyday drivers across 28 countries. the largest single-source driving dataset any open model has been trained on from scratch.
the receipts:
> 60%+ better image quality than the prior best (FID 2.72 vs 6.9)
> ~70% better video consistency (FVD 25.50 vs 82.8)
> 200+ models trained just to find the scaling rules
> final run landed within 3.6% of what they predicted before spending the compute
and the part that matters: this isn't a demo. a major automotive company used community-collected data as the complete foundation for a new open-source state of the art. that's real demand for a decentralized data network, not a pitch for one.
first Orbis 2, now VATIX 9B. community data is becoming actual infrastructure for physical AI.
How far can 5,500 hours of real-world driving data take you?
All the way to VATIX, a new open-source, state-of-the-art world model for driving video generation. 🚗
The first major result from the NATIX × @Valeo_Group collaboration is here!
SpaceXAI engineer, ex-Cursor:
"my most valuable Grok Bot writes zero code
it just remembers what the other 20 are doing, so nothing ever gets done twice
that one bot saved me 30 hours a week and a $9k hire"
in a 55-minute session, a SpaceXAI engineer wires that memory-keeper bot on top of a 20-bot team from scratch
this is the piece almost everyone skips, and it's why their agents rot by week three
watch it, then build the same memory into your team with the 12-step system in the article below