kentlyons

I can imagine. I did a bit of poking with blender and the api use got better there as well. For my use case I needed BREPs so stuck with freecad. I'm tempted one of these days where I have a pile of extra tokens to see about converting it to directly use the Open Cascade library since I just need to model and not the UI.

Claude Code is now WAY better at the freecad python api. My product (related to ceramics) does a whole bunch of custom 3d modeling and the new generation of LLMs can go in and make significant changes and basically get it right. They can also deal with a lot of annoying corner cases when you run into them.

This is very common in Europe. PhDs are sponsored by a given company

It just updated. Maybe affected by their own outage!

The actual rule for this part is farther down. Section 465.7b (p161 of the pdf). My reading is basically if the website is showing something that it looks like all of the reviews, then those can't be filtered in this way. But that seems to leave open cherry picking reviews - eg don't imply you're showing them all.

  ... receiving and displaying consumer reviews
  represent most or all the reviews submitted to the website or platform when reviews are being
  suppressed (i.e., not displayable) based upon their ratings or their negative sentiment...

I'm allergic to sawdust so went ceramics instead. Making things is awesome and it's always lots of fun with the anticipation of opening the kiln!

Yep - obviously corner cases are left up to the reader! The other way to think about this is to work backwards from your guess of a final price. So that would give you the size of each step. Likewise 20% is a pretty random number and can also be tuned. The key insight is the data and math should drive the decision.

Assuming you want to maximize income (and in turn, profit), this becomes a math problem based on data. Income = Price * Conversion Rate. So assuming a constant number of people coming in to your funnel (which should be the case since it is ~independent of price), you can keep increasing price (which likely decrease conversion rate, but not always) until Income goes down. To start, you don't know conversion rate, so set price very low (0 is a good start for a few reasons), and every N sales, increase it by 20% until income stops going up.

The city of San Jose is spread over a huge area (a good fraction of Santa Clara Valley aka Silicon Valley). The downtown area of San Jose which you might think of as a city is rather small.

And for small businesses, there is a massive wave of boomers retiring. Either they sell or go out of business. For the better ones, PE is buying and rolling them up.

Was management in place so when the founders exited the company would still grow? With lower valuations (1-3x), the purchaser is often buying a job in some way (either for themselves or needing to find an operator). At 7-10x multiple, the company is already has senior management in place so the new owners would expect continued growth without their own intervention.

There's a few factors. Are there that many, or is there just a lot of news? And related to a lot of news, there are a lot of orgs chasing AI whether they understand it or not. So to that end, as a founder, one could convince the naive (about AI) money to invest. Or more likely, the founders have a track record in some way, and then it is not much different than raising for a traditional startup.

I've thought a lot about these issues - I am actively working on creating new research and then commercializing it. However I think the incentives of investors (VCs and likely angels) are not aligned well with research development. As a result, I've landed on the bootstrap/self fund side of the argument, much like Midjourney. Find the low hanging fruit on the research side that can be monetized, and build off that.

Having worked at both. Hardware companies treat thier software like hardware. They spend a ton of time planning and iterate very slowly. It's very waterfall. I think one of the key advantages Tesla has is they are the opposite. They iterate on their hardware like software. They make hardware changes all the time so they can hill climb faster. And while Toyota invented TPS they seem to be stuck at some local maxima and the scope of changes they make with hardware are limited or slow compared to software.