John Resig (@jeresig) header
John Resig (@jeresig) profile picture

Creator of @jquery, Chief Software Architect at @khanacademy, Creator of https://xitter.gq/t.co/UcOO6CdK7c

Hudson Valley, NY
Joined February 2007
Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)Photo by John Resig (@jeresig)
Seeing this makes the end-state clearer to me: code that can't be read by humans, run on machines that humans can't access, via inaccessible APIs, trained on data that humans didn't create. LLMs are starting to get really good at manipulating mostly-opaque-to-humans code (see: their improved ability to reverse engineer machine code/assembly). Additionally, folks are accessing more-and-more of their LLMs in cloud-based environments without ever actually interacting with an LLM API (myself included). I think it's going to become increasingly likely that the end state is one in which you ask an agentic company to produce some result and it does it via [some system completely opaque to humans] and produces [some output in a format that humans are unable to interpret]. It makes me wonder if our current state of competing LLMs cross-training on each other will eventually become literally impossible (not just logistically so) when they become fully self-recursive and are producing output that even other LLMs can't understand, let alone humans. I think we're in something of a golden period here where humans are still capable of understanding how things work and how things are run (shell commands! python! node.js!) but it's going to quickly become irrelevant. We're already past the previous high-point where we were able to see reasoning traces for the top-level LLMs. I'm skeptical that open models will ever get to this point as the economics are working against them (I suspect that they will always need human-generated training materials in one form or another). But it does make me wonder if the upper-echelon companies will end up converging into this high-level "black box that produces outputs" singularity of sorts. This has been hitting home for me lately as I've been using Claude Projects in a cloud-based environment where I can't see transcripts, look at the machine it's running on, nor even know what commands it's running. I don't even know what languages its writing its code in. We're just a couple steps away from code being completely irrelevant.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧡 Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
42
36
10
282
48,191
My first time seeing halloween decorations based upon a Japanese print! Extremely cool - it depicts Princess Takiyasha summoning a skeleton spectre to frighten Oya Taro Mitsukuni at her father's ruined palace, famous Kuniyoshi print. reddit.com/r/ukiyoe/comments…
1
11
3,218
I love how good LLMs are at training other models. I used sol & astra to do annotations and classification on 10s of thousands of prints, then used opus-5.5 to manage the actual training, and now I have a fine-tuned detector and a classifier I can use to help categorize prints!
16
4,002
There's this one print series by Hiroshige & Kuniyoshi that has this huge title cartouche that causes my image similarity models to match every other print in the series. 15% of my DB is matches from this one series! I have to hardcode an ignore for this specific cartouche!
2
1
1
10
4,609
Was really hard to narrow this down. Many of these games led me to playing other similar games, which is why they're so important to me. Nowadays I'm mostly playing roguelikes, narrative games, and CRPGs. Not listed: Tetris-likes, Minecraft, and Platformers!
4
30
6,060
If you were confused by this announcement like I was (I already use Claude Code cloud a lot!) this is a credit to pay for some token use in the cloud. I spent 30mins trying to find the new pricing (that didn't exist). Cloud VMs won't cost anything extra (as has been the case).
Cloud sessions are officially available and out of research preview! They let you keep Claude Code working, even when your laptop is closed. Existing subscribers get a one-time credit to try them: $100 on Pro, $250 on Max.
5
1
1
16
8,188
gpt-6-luna (flex, low thinking) is phenomenal for text extraction. Half the cost of the next cheapest and just as accurate as gpt-5.6-luna. I switched from gemini-2.5-flash-lite when I started getting refusals for text extraction (?). Much better quality with luna in general.
6
5
80
9,100
Switched our internal @KhanAcademyEng PR review bot over from Opus 5 to Opus 5.5: Cost decreased by 50% Tool calls decreased by 34% Execution time decreased by 62% No decrease in quality.
10
6
1
179
12,720
I just had to explore using Jev with my existing Japanese print metadata extraction pipeline (where I use gpt-5.6 luna). Turns out that in some cases I could replace luna completely and save a bunch of money - in others I could augment what I had for higher quality!
6
1
1
41
7,896
I love giving Fable problems that I 100% would've never done otherwise, such as collecting Japanese prints out of static image auction catalogs written entirely in Japanese. Uses OpenCV to do segmenting, GPT-5.6 Luna to do text extraction + translation.
3
33
7,263
A little experimental game I'm making: carving Japanese-style woodblock prints! Wanted something chill, atmospheric, but most importantly: accurate to the actual experience of carving woodblocks! Still messing around with it but I've been having fun building it!
3
1
36
7,893
Incredible - thank you for the reminder, @HenriHelvetica! 20 years!!
πŸŽ‚ Many of those responsible are quite possibly no longer on this platform, but 20yrs ago TODAY the mythical @jquery launched v1.0. As usual, sharing a thread I posted after a short email exchange I had w/ @jeresig a few yrs back, ICYMI. xitter.gq/HenriHelvetica/status/…
19
51
2
488
38,477
Ripped Google Analytics out of ukiyo-e.org this past week and replaced it with a better, more detailed, home-grown setup that doesn't use any cookies! I didn't realize how much traffic I'm sending to museums - about 3,700 clicks per day!
8
2
46
8,901
"Impressions" is when someone sees an image inside a search result -> 5% conversion to a viewing a print image on my site -> 27% conversion clicking onto the final museum site. Averaging about 1,500 manual searches and about 200 search-by-image per day. Neat!
1
1
2,039
I've finally found a Wordle-like game that I'm good at! You have to associate historical artifacts with their origin and time of creation. Very fun! Anthropeum.com · Jul 31 2026 🟦🟩🟩🟨🟩🟩🟩🟦🟦🟦 82,910 · top 2% of players today!
2
2
16
7,746
The new Github Stacked PRs preview is incredible. Landing 5 stacked PRs directly to a merge queue all at once! A+++! This removes so much friction (and the gh cli tools + agent skill help a ton) github.github.com/gh-stack/
4
15
116
31,148
I have a physical object that I can take photos of from all angles and I want to turn it into a relatively low-poly 3d model for a game (mapping the photos onto the surface). I feel like there's gotta be some simple non-AI tool to handle this?
10
16
9,536
The Washington Monument on the Potomac River, Kawase Hasui, 1935 - depicting the cherry trees that were a gift from the Japanese government and planted in 1912.
12
33
4,996
My pie-in-the-sky requests for @CloudflareDev to overhaul corporate web dev: * Fully hosted private Git + CI + "Pull Requests" * Playwright end-to-end test running and result collection * Storybook hosting + shapshot comparisons * Private NPM package registry hosting
6
2
1
52
6,680
I feel like too many Github competitors try to focus on the public-facing part of Github. There are too many security problems there. So much can be gained by keeping things private and protected, especially with agents running in there.
1
7
2,248