Phillip Yan

industrials x AI. founder @MaingenInc (yc s26). semper quaerens cogito et facio. prev @Coinbase, @Columbia and @Oxford

[Phillip Yan @phillipyan]

Today we’re officially launching @MaingenInc on YC.

Frontier models have access to rich training environments for coding, finance, and legal, but not yet in the industrial economy.

Maingen builds simulations of real industrial operations, so agents can run the physical world.

[Brian Daley @brianpdaley]

Immortality won't come from brain uploading. It'll come from keeping the brain itself alive. Forever.

1. Keep the brain alive in a vat indefinitely. Vital fluids circulating, nothing that can kill it. Consciousness lives in the brain. Protect the brain, protect the self.
2. Engineer every sensory input. Dispensable robot bodies relay the real world. Or feed it a simulated paradise of your choosing.
3. Halt aging, cancer, and disease in brain tissue. Make the life-support resources abundant and unending.
4. Make the vats indestructible. Eventually off-planet, beyond war, terrorism, and disaster.
5. Engineer intergalactic travel for new resources. The only deadline left is the heat death of the universe. We'll have billions of years to solve that too.

Immortality and eternal heaven are achievable. We must engineer them.

[Phillip Yan @phillipyan]

@brianpdaley engineer the future you want to live in

[Phillip Yan @phillipyan]

@athreyaanand @hwintelligence congrats on the launch!

[Phillip Yan @phillipyan]

@kushagrchitkar Congrats on the launch!

[Daniel Rupawalla @danielrupawalla]

if i had to give one piece of advice to RL data vendors, it would be to focus on a single niche.

the biggest issue with naive vendors is that they don't see that the problems are different in each niche. there isn't a single blanket solution to scaling synth pipelines and they try to scale too quickly.

many core QA and design principles cross-over, but each domain has its own unique issues that must be hill-climbed. even domain-specific niches have bespoke issues. MLE and inference eng tasks are vastly different to enterprise SWE tasks. banking excel tasks are different from back-office or email tasks. there is SO much variance, each one indexes on a different axis of quality. please focus on quality, it makes all of our models collectively better.

[Phillip Yan @phillipyan]

finally got @phillipyan

[Brendan Ashworth @brendanashwrth]

150k+ civil engineers in the US manage $1T in infrastructure spend, yet only 0.8% of civil engineering work is done with AI.

We created SurveyorBench to evaluate LLMs on surveying workflows, a key part of nearly every civil engineering project (1/5)

[Phillip Yan @phillipyan]

@sumukhanadig8 @IgorZIJ so real

[Phillip Yan @phillipyan]

@IgorZIJ This is amazing, congrats!

[Phillip Yan @phillipyan]

@viktor_malstrom @sanskxr02 agreed

[Phillip Yan @phillipyan]

@sanskxr02 This is so real .

[Phillip Yan @phillipyan]

@GaddipatiHarsha looking good

[Phillip Yan @phillipyan]

@AnselDias @fdotinc super cool

[Phillip Yan @phillipyan]

@byvivarmaa @cinamarina @khaaloapp this is super cool

[Phillip Yan @phillipyan]

@CarlsonJustin @bryan_johnson this is impressive

[Phillip Yan @phillipyan]

@vikramnarra13 @bryan_johnson super cool

[Phillip Yan @phillipyan]

@sfxnz @bryan_johnson that's innovation

[Phillip Yan @phillipyan]

@candrewlee @bryan_johnson @midjourney agree

[Phillip Yan @phillipyan]

@vstalingrady @bryan_johnson wow crazy idea

[Phillip Yan @phillipyan]

@nickdamian0 @bryan_johnson @andromedasurg super cool, congrats