Twitxy Saved

Jamie Gull

Verified account This account is verified. Learn more

@jamiegull

Founder/GP Wave Function Ventures investing in early stage deep tech. Former SpaceX engineer on Falcon 9 reentry and Co-Founder/CEO Talyn Air, Y Combinator.
Los Angeles, CA wavefunction.vc Joined March 2009
3.4K Following 3K Followers 1.9K Posts

Latest posts 5

Pinned post

Jamie Gull Verified account @jamiegull
Stoked to get to finally announce Wave Function Ventures $15M Fund 1! Wave Function was created to partner with deep tech founders building hardware solutions to the world’s most important problems. I'm 9 investments in, with 15-20 left to go.
44 replies 19 reposts 355 likes 740K views View on X
Jamie Gull Verified account @jamiegull
Well, Chinese spies were not on my AI infrastructure bingo card.

The NVIDIA news is pretty wild, and it's a good reminder of just how quickly things can change when geopolitics gets involved.

I keep seeing startups where the entire go-to-market strategy seems to be selling to hyperscalers. That's a hell of a lot of eggs in one basket. You should absolutely sell into the AI boom and take advantage of the massive amount of capital being spent right now. Just make sure there are other customers for what you're building.

Between geopolitics, export restrictions, and changing spending priorities, there are a lot of things outside your control that could change where all that money goes. You don't want your entire business depending on a handful of buyers continuing to spend at this pace forever.

You can be completely right about where AI is going and still get killed in the transition.
3 replies 0 reposts 12 likes 707 views View on X
Jamie Gull Verified account @jamiegull
This is specialized robotics done right. Solving a hard problem, high willingness to pay, strong pain point.
Dustin Walper Verified account @DustinWalper 5d
Okay, so this is freaking awesome. Huge breakthrough.

This is what our autonomous shipbuilding process looks like for placing, fitting, tacking, and welding.

This is *one click* based on a 3D model. Our software does the rest:

* Find the correct part, pick, and place it to within ~1/32" of nominal, making complete contact with another part. No fixtures.
* Correct any sweep & bow in the metal to within 1/16" dimensional tolerances and tack it in place.
* Complete the full weld sequence, in this case a double continuous 3/16" fillet using metal cored wire.
* (Not shown) In-process laser scan of the weld for QC.

I've sped it up, but actual elapsed time for this operation is ~5:35. And it will get faster.

If we want to go fast, we need to think about cycle time across *the entire process*. Not just welding.

The steps we can do robotically take multiple people working much longer to complete in a shipyard - we're expecting to reduce cycle times by anywhere from 10-16x in realistic conditions.
View quoted post on X