freaky fish guy

hooked on something new, ex @waymo, @google

[freaky fish guy @samliu]

@juliarturc also,
* ai voice is still grating
* quality visualizations is still not solved imo
* youtube on walk is more reliable and low effort than say claude mobile app

[freaky fish guy @samliu]

@juliarturc the unknown unknown - a lot of people don't know what to ask about. subscribing to a technical explainer who is explaining the most recent / impactful things is a great way to build a list of questions, then go to AI to fill gaps

[Silen Naihin @silennai]

Excited to open source our world-model-harness!

`wmh` makes it easy to go from agent traces -> faithful replication of your production environment

Basically, an LLM pretends to be a Docker container but 5x faster

Below is a comparison running 8 SWE-bench tasks

[freaky fish guy @samliu]

The real nightmare is that fish stocks are basically inverse to inference markets

[alli @sonofalli]

husband couldn’t sleep last night because he had nightmares about how “inference markets will be an order of magnitude larger than anything we’ve ever seen before”

[freaky fish guy @samliu]

I didn’t read the other parts but fishing is truly best where others don’t go

This is not metaphorical

There be poison oak

Again not metaphorical you will itch for weeks

[Tim Ferriss @tferriss]

Ninety-nine percent of people in the world are convinced they are incapable of achieving great things, so they aim for the mediocre.

The level of competition is thus fiercest for “realistic” goals, paradoxically making them the most time- and energy-consuming.

If you are insecure, guess what? The rest of the world is, too.

Do not overestimate the competition and underestimate yourself. You are better than you think.

Unreasonable and unrealistic goals are easier to achieve for yet another reason.

Having an unusually large goal is an adrenaline infusion that provides the endurance to overcome the inevitable trials and tribulations that go along with any goal. Realistic goals, goals restricted to the average ambition level, are uninspiring and will only fuel you through the first or second problem, at which point you throw in the towel.

If the potential payoff is mediocre or average, so is your effort.

The fishing is best where the fewest go, and the collective insecurity of the world makes it easy for people to hit home runs while everyone else is aiming for base hits.

There is just less competition for bigger goals.

[freaky fish guy @samliu]

* camelcamelcamel but for https://t.co/ybXtX2fRDE

[freaky fish guy @samliu]

things I wish someone would make, june 2026 edition

* all-in-one solar charging board that supports parallel lifepo4 (3.2v) cells, 2-10w 5v solar panel inputs, and boost conversion. Tired of personal projects relying on short-lived lipo or crappy li-ion

[freaky fish guy @samliu]

* privacy respecting nanny software. i.e. "your child may be chatting with a stranger -- ask them?" or "your child seems to be getting bullied" vs "here's your child's text message and browser history"

don't wanna spy on my child, just wanna know they're OK

[roon @tszzl]

my waymo is conscious

[freaky fish guy @samliu]

@g1455mountain @kalomaze the representation can be better than stft on some axes, like for example if your task requires both phase and magnitude, the learned representation can colocate that info in a way that's more useful to the net. if the goal is isolating and transcribing speech, phase matters

[freaky fish guy @samliu]

@mattvanswol In HS we did the 90 A / 80 B / 70 C system, but I realized 89.5 rounded up to an A at my particular school

...which let me calculate a safe margin to fail by every finals season. tfw you write your AP US Hist essay about napoleon instead of us presidents

[freaky fish guy @samliu]

@arnie_hacker yes

[freaky fish guy @samliu]

@chanwoopark20 @spectate_or (Also relevant: https://t.co/ZxkbpNUMxM)

[freaky fish guy @samliu]

@chanwoopark20 @spectate_or All that said I believe in research automation as far as I think most of the "work" researchers are used to doing is grunt work and like every other discipline you can farm more and more out to AI and focus on strategic decisionmaking / hypothesis generation

[freaky fish guy @samliu]

@AaronBergman18 honest question, what's bad about using learnings from interp to build more capable models? from my perspective seems like an obvious direction all interp companies should ultimately take