Solo sailors became a semiprominent microniche on youtube during the pandemic, you can watch this sort of thing in incredible detail that way if you want. Sam Holmes did this trip (and many others) and has IMO the best channel about this kind of sailing. He has nerves of steel and is not necessarily to be emulated in all practices but a great watch. FWIW I'm a long-time recreational sailor but I've only done a few long passages and none of them solo.
Man, I really enjoyed Sam, until he started doing things that I thought were unnecessarily risky (or in some cases plain inconsiderate). That said, ballsy MF. Solo sailors remind me of free solo climbers. Their risk assessment/acceptance is so far off normal it's almost alien.
I haven't watched all his videos especially the more recent ones but there have definitely been some wtf moments for me too. And yeah he's not ignorant at all it's pretty funny to watch him verbally run through an accurate & extremely alarming assessment of the dangers of something and then go like eh it'll probably be fine.
"Well we're going to take an unplotted shortcut here through the uh, <<checking notes>> graveyard of the atlantic, according to the charts it's 7 feet deep and we have a 6.5 foot draft so fingers crossed the charts are up to date" type shit.
IIRC he has some very serious chronic health stuff and that may have something to do with it. I don't love the risky sails when he has a passenger though. But also who the fuck am I, a weekend lake sailor, to judge. At this point he's got to be among the most experienced solo sailors on the planet. I feel a similar way about slocum or the pardeys and they are more his peers than I am.
The interesting question to me is how far these reasoning models can be scaled. With another 12 months of compute scaling (for synthetic data generation and RL) how good will these models be at coding? I talked with Finbarr Timbers (ex-DeepMind) yesterday about this and his take is that we'll hit diminishing returns – not because we can't make models more powerful, but because we're approaching diminishing returns in areas that matter to users and that AI models may be nearing a plateau where capability gains matter less than UX.
I think in a lot of ways we are already there. Users are clearly already having difficulty seeing which model is better or if new models are improving over old models. People go back to the same gotcha questions and get different answers based on the random seed. Even the benchmarks are getting very saturated.
These models already do an excellent job with your homework, your corporate PowerPoints and your idle questions. At some point only experts would be able to decide if one response was really better than another.
Our biggest challenge is going to be finding problem domains with low performance that we can still scale up to human performance. And those will be so niche that no one will care.
Agents on the other hand still have a lot of potential. If you can get a model to stay on task with long context and remain grounded then you can start firing your staff.
For people like the OP seeking high quality resources with an emphasis on the networking/sales/marketing skills required to succeed as a tech/dev/software consultant, I'd recommend starting with "Getting Starting in Consulting" by Weiss.
Then if you'd like more after finishing that, others to check out are:
Also, I'd like to talk with anyone actively trying to shift from being a software engineer full-time to being a software consultant. Or anyone trying to grow a new consultancy. I'm exploring starting a business that makes it easier for engineers to successfully transition to consulting. I'm just at the early pre-product stage, and am still trying to better understand the new or prospective software consultant perspective. I seem to get along well with people from HN so if anyone here is happy to help and jump on the phone with me, that'd be hugely appreciated! My email is in my profile.
> Eh, I still think the best thing for the gov't to do in regards to SS numbers is just publish them all in an open database accessible to anyone.
They basically are available to any business that needs them. Just takes a few months to go through the compliance process for an identity verification company. See https://cognitohq.com/docs
Regarding it being a password, yeah that is pretty bad. Some friends of mine are working on Bloom (https://hellobloom.io/) which is trying interesting things that hook into the real world (necessary) while also trying to create a better system eventually. Basically, an ID you control with a private key is approved by identity verification services. Your friends basically vouch for the ID being correct. Your identity is tied to that ID. If you lose the private key then you have your friends invalidate the old ID and transfer their vouch to your new ID. Way better than just a secret number.
I suspect just requiring people seeking credit to just have to come personally to a bank with a photo-ID, a pay-stub and have to confirm they have access to the postal address they have on their ID would solve it. This wouldn't by any means be foolproof (after all, half the highschool students in the US manage to get passable fake IDs), but the fact that it would increase the effort to open lines of credit, slow down the rate at which you could open new lines and increase the risk of getting arrested when you tried would, I think, shift the risk and reward balance on most scams to the extent that it wouldn't be worth it to the scammers anymore
Don't let the homepage confuse you -- after playing around for a bit, it seems like this is basically a nicer (IMO) and definitely simpler (less featured) version of Squarespace, also currently free. And looks like it works with custom domains too.
