Interview with Sulaiman Ghori, Member of Technical Staff at xAI. XAI will emulate millions of humans via the distributed Tesla chips. HW4 good enough
AI and XAI are hardware constrained and the biggest edge for XAI is speed of hardware deployment.
Predicting future bottlenecks
Elon excels at forecasting bottlenecks months and years ahead and working backwards. XAI Teams adopt this by focusing on core metrics (financial/physical). Eliminate perceived software limitations (latency, overhead) for 2–8× gains.
Macrohard model iterations happen daily and multiple times a day (even from pre-training). Hardware racks start training within hours (sometimes same day/few hours) of setup—vs. days/weeks elsewhere. Supercompute team removes typical barriers.
He joined XAI after startup attempts and outreach from Greg Yang.
Bootstrapping off the Tesla network
Potential to deploy millions of human emulators cheaply on idle Tesla vehicles (Hardware 4, networking/cooling/power already present). There are about 3 million HW4 cars in the USA. Pay owners to lease compute time—more capital-efficient than AWS/Oracle/NVIDIA hardware. Enables massive scale without new buildouts.
What is Macrohard
Digital equivalent of Optimus: emulate any keyboard/mouse/screen-based human digital task at fraction of cost + 24/7 uptime. No software adoption needed. Rollout will be slow at first, then rapid scaling (1,000 → 1M emulators not infrastructure-limited).
How Elon Deals with Problems or Fires
Immediate action lke phone calls to vendors for patches next day, side-by-side fixes.
Blockers resolved in hours/days.
Frequently asks “How can I help make this faster?” at end of meetings.
What it’s like working at xAI
High freedom/responsibility. Jump projects, own components quickly. Overlap/flow between 2–3 projects. Fast cycles (24-hour Imagine rollout iterations). Intense pushes (weekends, long office stretches). Talent density requires hard work to keep up.
Using 80 mobile generators + battery packs to balance load at their data centers
Seamless switch to 80+ mobile generators during high municipal load to avoid grid impact. Batteries enable fast megawatt-scale up/down response (vs. slow physical generators). Layered: capacitors → batteries → generators → grid.
How they built Colossus in 122 days
Temporary/short-term land lease via carnival exception for fast permitting. Assumed resources available; high utilization with many parallel runs by small teams.
Work backwards & figure out the highest leverage thing you can be doing
Start from revenue targets (e$10–100B), reverse-engineer physical/software needs. Highest leverage first; physical requirements last.
How xAI hires
Hackathons (high ROI even for few hires). High value-per-commit (~$2.5M per main repo commit). Seek simple solutions.
Challenge assumptions/requirements
– include deliberate errors/impossibles to test pushback.
Everyone’s an engineer
Flat structure. Managers code. Even sales/enterprise teams train models. Everyone contributes technically.
Testing human emulators as employees
Internal virtual “employees” → funny incidents (invites to non-existent desks). Generalization better than expected (handles untrained tasks flawlessly).
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
