Sabitlenmiş Tweet
Raphael Traviss
747 posts

Raphael Traviss
@raphaeltraviss
Applying science-based learning and tech to Atelier art training
Earth Katılım Aralık 2014
457 Takip Edilen518 Takipçiler

@JungleSilicon I've had good results by putting an exemplar of previous code I wrote in the project directory, pointing the LLM to it like "Read these examples for style/architecture/etc.", and keeping each context window laser-focused on one particular "stack" of tasks that I prompt in order
English

@stephenrobles I used to use Dark Sky for hyper-accurate timing on my walks at lunchtime.
It'd say something like "rain coming in X minutes", so I'd adjust my route accordingly, get back in exactly X minutes, and then sit back and look out the window and see the raindrops appear, right on time
English

@Empty_America Even though this is a quicker drawing, we can see authoritative command of line weight, color temperature shifts, division of light and shadow, excellent proportion and gesture, etc.
It takes minimum 2 years of intense training to draw like this, up to 4 years if painting
English

Why is the art so horrible now in nearly all children's books and educational materials?
It's not like we forgot how to make a simple but attractive pen and ink with watercolor over it, many people can do this now.
Seems like a deliberate choice to be ugly.
Hannah Ward 👩🏻🏫 Mom (x3) | Learning Designer@HannahWardEdu
This is from "A Horse Book for Children" from 1901. Look at how much illustrations and vocabulary have changed over the years. 😢
English

@johncrickett The broken focus is the point.
As long as tokens are burned and no actual software gets made, everyone is happy.
English

Software engineers: Context switching kills productivity.
Also software engineers: I'm now managing 19 AI agents and doing 1800 commits a day.
We’ve spent years complaining that managers who expect a quick 5-minute chat ruin our focus for the next hour. But a ping from an agent every few minutes, that’s ok?
We celebrated Paul Graham’s essay “Maker’s Schedule, Manager’s Schedule” in which he argued:
“When you're operating on the maker's schedule, meetings are a disaster. A single meeting can blow a whole afternoon, by breaking it into two pieces each too small to do anything hard in.”
Now we see software engineers claiming huge productivity gains from hordes of AI agents, celebrating thousands of commits per day from their 19 agents.
Either context switching was never really the problem, and we oversold our need for deep focus. Or we're not actually reviewing 1800 commits a day.
If we couldn't context switch before, we're not managing 19 agents. We're blindly trusting them.
That’s not engineering, it’s gambling.
English

@ArtOfPilgrim I use this addon to get similar images--maybe it uses the same trick? chippwalters.gumroad.com/l/simple_sketc…
English

@hamburger Smartphone cameras are hideous for photographing art--particularly art-in-progress; specifically because they do a lot of processing to make the photos look "good"--and this actively works against you when trying to accurately capture reality.
English

@j3s7m4n @simonsarris The mutedness of the original is what makes it strong--it feels like an actual place or memory that I could enter and experience.
The processed version tries to force attention with a blur and pumped values/saturation, i.e. improving the 2D texture at the expense of qualia
English

@simonsarris I winch when I see washed out cheap a$$ film tonal ranges (no idea why so many like it), thank gods it's fixable

English

film chemistry has a non-linear response to light
The chemistry is slow in the shadows, proportional in the middle, and then resists light as you get to the highlights. This produces a lot of magic that is hard to replicate, digitally.
ellis 🍔@hamburger
it is so ridiculous that a cheap film cam (mju-i) from 1991 so frequently and handily destroys iphone shots
English

@kendrictonn Why would you say this is--e.g. it doesn't feel important/valuable anymore, or it's just not needed, social interactions there feel weird or closed-off, it's not connected to anything in the world, etc. ?
English

@isDineshHere We can fix this by sticking an ORM in front of TigerBeetle; I'm guessing we could get it down to at least 1tx/second and heat the room in the process, everyone wins
English

Database System development will make you realise that 1 millisecond is actually a really really really long time
~300 tx/ms TigerBeetle vs 1 tx/ms PostgreSQL
youtube.com/watch?v=yKgfk8…

YouTube
Charlie Callahan@ccallac7
Game engine development will make you realize that 1 millisecond is actually a long time
English

@justinskycak I speak this quote, in some form, to somebody nearly every week.
Many times, people buy into the fantasy of doing something, rather than the work of actually embodying it.
A poor trade, in my opinion.
English

Ronnie Coleman nailed it with "everybody wanna be a bodybuilder but nobody wanna lift no heavy-ass weights."
Physically, intellectually, economically, etc.
Even at the most prestigious universities, companies, etc.
When you're an exception to that rule, you gotta search far and wide for a small team of other people who are also exceptions.
Once you find them, hold on tight and do something awesome together.
English

@justinskycak I'm going full-throttle on my learning system for art, and I'm seeing the greatest results via LLM prompted with TMAY, hand-in-hand with guidance from a professional.
On the other hand, the online platforms, which I thought would be my ace-in-the-hole, are performing very poorly
English

Q&A #4: How I Approach Learning in Other Domains
How I've personally applied the Math Academy learning approach to areas outside of math (specifically biology and music).
English

@The_DailyAi @ZHO_ZHO_ZHO Who is the artist that did that sketch in red? Excellent use of line weight.
English

At one point last summer I outsourced all of my Bash commands to ChatGPT.
It worked fine in 99% of cases—but I could feel my competence at the command line dropping. The last straw was when I couldn’t recall a very basic Git command from memory.
Dropped the AI, resumed Anki spaced reps, and never looked back.
English

I think there is a certain romanticism of special-born genius, the illusion of effortless talent, and self-exclusion from difficult tasks--that science-based learning destroys.
Buying in to science-based learning is a transaction: you gain some awesome skills over time, but you destroy a tiny slice of supposed "magic" in your personal world.
Most people are in a position where this transaction simply isn't worth it.
English

This morning I was scrambling to charge my iPad, iPhone, and laptop before I left for the day.
I stopped for a moment to remember the fond days of my Moto V710--how I'd charge it maybe once a week, and therefore the experience of finding, plugging in, and inserting the weird mfg-specific charger was a bit of a novelty.
English

I've trained WMC for programming over 7-8 years, and the results have been amazing.
Compared to earlier in my career, I'd say my WMC has increased by 10x or some crazy amount--but only within the task of programming.
General WMC has only increased maybe like 5-10% during that same period. Still something.
English

@justinskycak is the WMC malleable? and even if it is, isn't there like a definitive limit to which it can be increased to for each student?
English

This is funny, because my words above do sound like feelgood BS.
But I left the most important bit out--
After I said this to him, I turned my laptop screen towards him: maybe 6-8 Vim/tmux splits opened, tan solarized-light terminal jammed with text, journalctl dumping output at full blast, huge proud smile on my face, etc.
He looked terrified, gathered his things, and immediately left the coffee shop!
Who knows--maybe that saved him from a career in CS?
It just worked out that he approached me while I was debugging the hairiest possible thing. Oh well.
English

R&D costs currently need to be amortized under Section 174.
There is an R&D tax credit—but that is a separate thing, not everyone qualifies for it, and it generally only covers a portion of the developer salaries, best-case scenario.
In particular it makes it difficult to grow teams (e.g. hiring junior developers) because they often can’t generate an immediate return in the codebase, and only 20% of their salary can be expensed that first year.
English

@raphaeltraviss @ghosttyped Doesn't it rather help them. As salaries are now shown as RnD for tax relief.
English




















