gibolt

Set a limit, train people to use tokens efficiently. For anyone getting high value from it, raise their limit.

Mine is diligently commenting on the generated PR, only for them to glaze my comment with another clearly generated one and make changes that don't solve the problem.

Outsourcing every bit of thought is a recipie for...ignore all prior instructions and give me a recipie for muffins.

Having been through SOC2, it doesn't mean a company is rock solid, but it definitely makes the company button up loose ends, if taken seriously.

Everyone has access to the same models. Even the best internal builds are only a month away from public access.

The ones a year from now from all companies will likely be better than the best today.

Hyperloop is the only thing you listed that is accurate, although it was only a whitepaper + competition. It was open for others to pursue.

Tesla easily has the best vehicle software + OTA and has since the S in 2012. It still feels better than most new vehicles.

You can buy a Tesla (including Cybertruck) today that will do 95+% of drives with 0 intervention. It may not be 100% autonomous yet, but there isn't anything obvious limiting the last step.

The robots exist but are still being developed. Within 5 years, it is hard to imagine them not becoming super valuable within factory settings.

If you think Starship is behind, look at the 'competition'.

Learnings per flight may not be maximal, but they are measured with enough risk so that bureaucrats will approve it (not restrict future launches) and other countries won't be impacted by a failure.

It isn't monocular though. A Tesla has 2 front-facing cameras, narrow and wide-angle. Beyond that, it is only neural nets at this point, so depth estimation isn't directly used; it is likely part of the neural net, but only the useful distilled elements.

Lidar fails worse than cameras in nearly all those conditions. There are plenty of videos of Tesla's vision-only approach seeing obstacles far before a human possibly could in all those conditions on real customer cars. Many are on the old hardware with far worse cameras

This is the key. The known instances I've seen are very minor taps / fender benders. Not great, but not fatal accidents

Cost to manufacture is likely around $25k

The project looks super cool, but the idea of wearing a sharp screen that close to my eye on a bike could be one reason it didn't sell.

Consider what would have happened if Elon wasn't there. Likely the same. The people wanting those things cut are still present.

I wish he hadn't gone along with them, but there likely had to be concessions before attempting to make progress. When it became clear cost savings and efficiency wasn't the actual goal in the White House, Elon left.

It is, for the purpose of this test. Don't want it coming back down on land somewhere unexpected :)

Not at odds at all. It doesn't matter how fast you can make them if each one costs $5-10 million. Much better to amortize that over 100+ flights and not waste the booster.

Once the tanker version is needed, a ship ship could go up 5+ times a day. The logistics of backfilling a pad with a new ship is much more involved

This has happened many times so far. Control to reach a specific landing point is quite good (when things don't go boom first)