C41Shawn

doing some major rebuilding, you'll like it. c41shawn.com

@_love_be_ing_ @domeniquedumont Damn, give that knee time to hopefully heal correctly though. One of my knees was never 100% again after messing it up like 25 years ago. I took the brace off too early and stuck the healing at 90% or something. Live n learn.

sans cesse, mon cheri - domenique dumont @domeniquedumont 43 yo 44 in may 12, joyous gratitude torn acl and small knee fractures on mend the past many months, God just been walking me around and dancing it off, how i kerri strug injured the knee in the first place, joy innit cant wait to freestyle more to this song and newly discovered artist of heart

@batzdu @killianmoore_ Badass.

Don’t go clogging it up with 500 dumb skills from GitHub. Make a few levels down a Claude folder and put in there all your skills memory preferences and a Claude.md file. Create handoff file at session end/big tasks, include milestones fails and wins and learnings. Auto read at session start (then prune/limit/extend). Give it a browser to use. Use plan mode (shift tabx2 in Code). Before big actions make it verify. Ruthlessly write rules upon errors. Careful with your keys! And lastly, if it sings your praises but doesn’t call out your bullshit, write a rule 🛠️

If you are an artist my number one tip for the week is pay $20 for a basic Claude account. Install the desktop app on your computer. Start a project in cowork to audit yourself. Tell it to check all your twitter posts over the last 3-6 months, your DM's, emails, whatever. Share all your art with it, how you released it, when, etc. tell it to dive into the metadata to learn. Explain to it how the market was hot 4 years ago and sucks now to give it context. Ask it to analyze what you did that worked, what didn't work, who are your biggest champions. Ask it to explore your peers who are having the most success and ask what they are doing that you are not doing or could do better. Just keep going deeper. Keep teaching it. Tell it to ask you questions about things it may not understand. Keep giving it more and more context. If it says you are maxing out on everything and kicking ass on every marketing strategy and if your art still hasn't sold in 2 years, then it might be time to admit your art isn't desirable to the masses or you didn't network well enough or something and you should be honest with yourself about that too. But for $20 you might learn a lot. Also no matter what you conclude please go easy on yourself too cause the market is absolute shit atm. The key though is do it in a project so claude doesn't forget things you tell it as you dive deeper and deeper. Best wishes to all you artists. I hope that helps someone. 🥂

@_love_be_ing_ Gm G. It’s somehow both corporeal and intangible.

@C41Shawn gm shawn, yes !

I’m telling my grandkids this was the first ever AI music video. youtu.be/5mLjZRAXRRA?is…

I love film so much, I can't comprehend a life without it. I want you to feel that way, too. Six prints available soon, three open and three limited. Representing a very broad range of my recent favorites, and one all-time favorite. [detail: la nuit américaine]

@laurentCastell Happy birthday 📸

TODAY'S MY BIRTHDAY +1. STILL HERE. +1. STILL BUILDING. ANOTHER BLOCK VALIDATED. BUILT IN SILENCE. SHAPED BY TIME BECOMING THE WORK ITSELF.

Me. But using my forehead instead of a hammer.

@NftSariyildiz Thank you, kind sir.

@C41Shawn awesome

This is Los Angeles on 135 film.

Find some opportunities for yourself

@SupportJburn Somewhere between a few and a handful.

How many active artists are still in this space? How many active artists are still in this space that don't rely heavily on AI?

@5tr4n0 whoah an old classic one!

may the snow keep falling

this is rad! I followed your framework (roughly) and it's already pretty decent after just two test batches of about 100 photos each. blind so far, just with existing data and dates and such. but i'm thinking of trying a local agent system for the visual part. will be slow, but might be better that way for me. i'll post what happens.

the initial script was pretty good! some testing and tuning (basically just giving the AI my feedback with examples). the tokens for claude's vision API get used up SUPER fast. i think the most i've been able to process in a 24 hour period was only about 100. thats why i wrote the script in the first place was to try to bring the number of files i actually run on AI down. i'm planning a second pass of the script on the folder of selected works, too, since this first test batch of 300k frames will get cut down to 3k frames, so a second pass will cut it down to 300 frames. but i'm still tuning. mainly i give it false negative and false positive examples and some descriptive feedback and it tweaks the script. i had been testing on batches of 5000, but now i'm trying it out on the big batch of 300k. then i'll tune it some more and run another 300k, and do that a few times. because of the limits on the vision API, i may just do a human curation pass, then let Claude choose which pieces from that curation will actually get published

I'm teaching AI to curate my art, not create it. I have 20 years of raw VHS glitch footage sitting on hard drives. Hundreds of thousands of frames. Somewhere in there are incredible stills — moments where a recognizable subject is visible through the glitch artifacts. Finding them manually would take months. So I'm building a two-phase curation pipeline with Claude Code. Phase 1 is a local Python script that scores every frame on structure and glitch intensity — Claude wrote it, I tune it by reviewing the results and giving feedback. It processes 30 images per second at zero cost and eliminates 90-97% of frames as uninteresting. Phase 2 sends only the survivors to Claude's vision model for final scoring. The key breakthrough was multiplicative scoring: a frame needs BOTH a recognizable subject AND strong glitch artifacts to pass. Took 11 iterations to get here. Each round I'd find a new failure mode — scanlines fooling the filter, dark atmospheric frames getting rejected, abstract color blobs sneaking through — and we'd add a new metric to catch it. Currently processing 300k frames across multiple sources. Targeting over a million for calibration. The frames that survive are going to become a new collection.

inspired by @maxcapacity - i now have the initial build of a bespoke photo naming, tagging, organizing script. python built with claude. next step is infinite.

This is the type of AI use I love seeing. So much cool shit being made right now all over the place.

Incorrect edits of photos I took on film. Various locations in Sweden.

@seanmundyphoto Doh!

@C41Shawn This one is actually made in Blender! Thank you for sharing though 🔥🙏

Photography is a broad canvas.