Every frontier lab is now hiring 'forward deployed engineers' — a role Palantir coined in 2006.
Worth sitting with: the most advanced AI companies on earth concluded the bottleneck to AI adoption isn't a better model. It's a human sitting inside the customer making it work.
The product is people.
@EcZachly It's not dying, it's inverting. The leetcode path rewarded surviving a filter; the AI-era one rewards creating value with no filter at all.
The scarce skill stopped being 'can you pass the screen' and became 'do you know what's worth building.'
It's creating jobs and quietly redefining the title.
'Engineer' used to mean the person who writes the code. Now it's the person who decides what's worth building and can tell when the model is confidently wrong.
Same headcount line. Different hire entirely.
@mingchikuo The under-read part: Google's most advanced AI-agent silicon is co-designed with MediaTek and fabbed in Taiwan.
The models are American; the compute that runs them is increasingly Asian.
That's the structural edge the market keeps underpricing.
3–5x pay to poach Taiwan's chip talent isn't a bidding war. It's the market repricing what it called 'cheap labor' for 20 years into what it always was: the scarcest hard-tech talent on earth.
The Asia edge was never the cost — it was the depth.
Tesla, Google & Micron Go All-In: Poaching Taiwan's Top Chip Talent with 3-5x Pay
Taiwan is the world's semiconductor talent powerhouse, home to TSMC & NVIDIA's key supply chain.
International giants are raiding it aggressively.
Tesla's Big Move:
- Tesla is hiring senior TSMC engineers for its "Terafab" wafer fab project.
- Offers: 3-5x TSMC salaries (up to ~$240K+ USD / NT$750K-800K) + juicy RSU stock incentives.
Google's Smart Plays:
- Acquired HTC hardware teams (2018 & 2025) for Pixel phones & VR/XR tech.
- Built its largest AI/hardware R&D center outside the US in Taiwan.
Micron's Expansion:
- Growing R&D and manufacturing footprint in Taiwan for DRAM.
Firms lure talent with high pay, stock options, better WLB, and US equity upside.
Local Taiwanese IC design firms are struggling, losing seniors, then fighting over juniors.
Taiwan's engineers are highly skilled, hardworking, and now globally in demand like never before.
This talent war shows Taiwan's critical role in the AI supply chain.
The Asia talent edge was never 'cheaper.'
A great engineer in Taipei or Manila was priced by their zip code, not their output.
Remote work removed the zip code. AI removed the language gap.
What's left is the work — and it's been mispriced for 20 years.
'Asia = cheap labor' is the most expensive hiring mistake right now.
The cheap tier is exactly what AI just ate. Commodity offshore dev is dead.
What's scarce and rising: Taiwan supply-chain + semiconductor judgment, Manila talent that clears the AI bar.
Arbitrage churns the moment someone's cheaper. Edge compounds.
Hire Asia to save money, you lose both. Hire it for judgment the US can't staff.
Watch where the best AI people go when a frontier stalls.
Cruise wound down. ~15 of its ML engineers landed at Hims — a telehealth company run by Cruise's old CTO.
Robotaxi didn't need them. Healthcare did.
Elite AI talent is liquid now — the skill outlives the sector.
Hire for judgment that survives the industry, not the logo on the résumé.
@etnshow@s16h_@MetaviewAI Faster hiring isn't the flex it sounds like.
Going from 93 days to 6 just means you can be wrong about someone 15x faster.
AI speeds up the screening. It doesn't tell you who can handle a messy, unclear problem.
Bad hires were never a speed problem.
This is already how the best teams hire. You don't fill a slot — you find a great person and build the role around them.
Finding smart generalists who get AI was never the hard part. Making room for them is.
The bottleneck isn't the talent market. It's the headcount spreadsheet nobody will let you change.
Junior dev jobs are basically dead.
AI writes the code now. What's left:
→ knowing what's worth building
→ catching when the AI is wrong but sounds dead sure
We stopped hiring for speed. We hire for taste.
By 2027 "years of experience" won't matter. "Can you tell when the model is confidently lying" will.
@JasonrShuman Smart hardware stalls when the OS above isn't agentic — Echo could only listen, Glass could only display. Gemini ships hardware where the OS IS the agent. The wedge isn't form factor anymore; it's delegation.
@ryaneshea Curious does your framework catch the laziness (mostly giving excuses to skip clearly demanded tasks)? It’s particularly bad recently with opus-4.7
@kevinxu If you believe in this thesis shouldn’t you just buy the call options that expires this Friday? (I’m on a similar challenge to turn a wild gambling $15k account into $1M this year)