Technology just crossed the line where you can't tell a robot from a real woman even up close
Liang Wei, 34. Former Tesla engineer. Quit because he was tired of building a robot that was still learning not to fall over.
Took 11 people to Shenzhen and set one goal: make a robot face indistinguishable from a living one.
Tesla spends billions teaching Optimus to walk straight. Liang spent 2 years making a robot smile like a human.
9 months just on the smile. Over 800 iterations of the silicone shell. Each one filmed, compared to a real face, micro-differences found, rebuilt from scratch.
At the expo people walked up and for the first few seconds had no idea they were looking at a machine. Some said hello and tried to start a conversation.
Now he's sitting on $12 million in contracts from hotel chains and shopping centers that want these in their lobbies.
In a couple of years you'll walk into a store and won't know who's serving you. Does that excite you or terrify you
OpenAI just showed a system called dots that stops answering questions and starts marking where the answer would have to come from.
It returns no text. It returns points on your own data, each one a place the claim is either supported or is standing on nothing.
> SPREAD - the question broken into the individual claims an answer would need
> ANCHOR - each claim pinned to the exact span in your documents that carries it
> VOID - claims with no span behind them left empty instead of being filled in
> LINK - anchors connected where one claim depends on another holding first
> READ - the finished map returned as points, never as a paragraph
A normal answer hides its own gaps. The sentences are smooth, and the part with nothing underneath reads identically to the rest.
Dots refuses that trade. An unsupported claim shows up as an empty point, impossible to skim past.
The dependency links are the sharp part. One hollow anchor exposes every claim resting on top of it at once.
That turns a wrong answer into a located problem. You see which document is missing, not just that something feels off.
Fluency has been doing the convincing in every tool shipped so far. Here it has nothing to hide behind.
The dot format and how the voids are scored are laid out in the breakdown below.
holy sh*t, someone just turned an entire company into 76 Grok Bots
every builder can steal the whole stack
372 skills across engineering, product, marketing, sales, finance, legal, HR, data and the C-suite
the patterns:
→ route every request through a chief of staff bot that picks the department for you
→ ship code through a reviewer, a dependency auditor and a security scan before it merges
→ triage NDAs and vendor contracts before legal ever opens the file
→ build the board deck from the numbers the CFO bot already closed
→ run a scenario war room on a decision before you commit the budget
→ score a hire on people analytics instead of a gut call
→ turn a roadmap into specs, tickets and an A/B test plan in one pass
→ log every decision so next quarter knows why you chose it
→ keep one company OS doc that every bot reads before it acts
the idea:
chief of staff routes → 75 specialists do the work
one company bot isn't one prompt
it's 372 small skills with one job each, split across 76 Grok Bots someone already built, tested, and shipped
watch it, steal what you need, then read the full 12-step roadmap below ⭣
My entire payroll is $80 a month.
A competitor's is closer to $80,000.
And we ship at the same speed.
Theirs buys five people, a manager, and a calendar full of meetings.
Mine buys four agents that run in parallel while I approve — no salaries, no sick days, no notice.
The gap isn't talent anymore. It's overhead.
Stop competing on headcount.
Start competing on how little of it you need.