abraham.py

Building KnightVision. Computer Vision for physical ChessBoards @Jousting_Knight. Transforming OTB Chess for 600 million+ Chess Players.

[abraham.py @def__abraham__]

GPT-5.6 Sol just one-shotted this Chess app that can record your OTB games in real time and it’s insanely good 🤯🤯🤯

[Gideon. @black_boi__]

Nepotism that transcends death. Legendary.

[Jennifer Shahade @JenShahade]

Fight for your own mind!! ♟️📚

[abraham.py @def__abraham__]

The eval bar is chess's scoreboard, but it only exists for boards wired with sensors.

I built KnightVision so a phone camera can track any OTB game in real time.

Every board in the hall gets instant context.

[Daniel Rensch @DanielRensch]

How would YOU take chess broadcasts to the next level? 📈

[abraham.py @def__abraham__]

@taylorhrsn @knightcapchess Awesome! I'd love to collaborate. If people bring their tripods, they can set their phones up to record their games with KnightVision so they could replay, share, or livestream. We could even have a website that shows some of the games played!

[Taylor Harrison @taylorhrsn]

international chess day is in a week!

last year we took over The Well in toronto with one long table and just let people play. every board was full in 5 minutes. turned into one of my favourite things we've done with @knightcapchess

doing it again this year.

[abraham.py @def__abraham__]

@OkezieEbi Yeah 😂

[abraham.py @def__abraham__]

I don’t care about 2027.

Where is my M5 Ultra 512GB RAM that was supposed to be out this year?

[AppleTrack @appltrack]

Apple's M7 Ultra chip coming in 2029 is rumored to support 1.5TB of RAM.

This would make the processor much more capable for on-device AI.

[abraham.py @def__abraham__]

@rajachess KnightVision:

https://t.co/J9QdnhQf1P

[abraham.py @def__abraham__]

People already play most of their chess online. Every game saved, every blunder logged, every opening tracked. Whether we like it or not, the digital game keeps getting more convenient than the real one.

But there are still a few places where online chess can’t compete.

One of them is over the board.

No app can replicate the tension of a tournament hall. No server knows what it feels like to shake your opponent’s hand, flag in a won position, or grind out a draw with your clock at ten seconds. Chess at its best is still played on real boards, between real people.

Ironically, those are the only games that don’t get saved.

Your online games archive themselves. Your OTB games, the ones you actually sweated over, live and die on a paper scoresheet. Most players never analyze them at all.

That’s exactly why I built KnightVision. Point your phone at the board and it reads the position. Snap your scoresheet and it becomes a full game you can replay and analyze. No electronic board, no manual entry, no losing the games that matter most.

Download KnightVision.

[Anish Giri @anishgiri]

People already prefer AI-written posts over many human ones.

Soon, many people will prefer AI tutors, AI assistants, AI friends, and even AI nurses. Whether we like it or not, AI keeps getting better at more and more of the things we once thought only humans could do.

But there are still a few places where humans remain irreplaceable.

One of them is over the chessboard.

No engine can experience the tension of defending a worse position for six hours. No AI knows what it feels like to blunder in time trouble, recover from a devastating loss, or win a tournament with one final game. Chess is still about people competing with people.

Ironically, the best way to prepare for that very human experience is with AI.

That’s exactly why we built ChessMonitor. It combines the latest engines, opening databases, and opponent preparation tools to help you spend less time searching and more time actually preparing.

The games are still played by humans. Make sure you’re the better prepared one.

Download the ChessMonitor app.

[ali @waterloo_intern]

it took 1 intern 3 months of continuous work, but eventually, a quantization method that beat every other algo in the market, including @nvidia's official modelopt

to explain why this matters, i ask for exactly 69 seconds of your attention (275 words @ avg reading speed of 238 wpm):

frontier models (like glm52) are huge (~0.8T params). as released, each parameter takes 2 bytes (bf16), so overall size is about 1.6 tb

a b200 has 180gb of memory. a node of 8 gives you 1.44 tb, barely fits weights, much less activations / kv cache

must quantize the model (reduce the size of each individual parameters) to serve. fp8 quantization means each parameter takes 1 byte (fits in 0.8 tb), fp4 takes 1/2 a byte (fits in 0.4 tb)

cutting the model to a quarter its original size is necessary for it to run a) cheap b) fast, and every lab serving models does this.

but, quantization lobotomizes the model if not done correctly (this is why you see people complain about @AnthropicAI nerfing claude or @OpenAI nerfing codex)

there are currently several algorithms (like Nvidia's official model-opt) that attempt to figure how to quantize a model with the least amount of damage.

they find the redundant layers that can be slashed, and sensitive/important layers that need to stay in full-precision.

these algo's have two drawbacks:
1) they take a long time to run
2) they quite often result in a sub-optimal configuration

for the past 3 months, a research (and, as always, waterloo) intern on our model perf team (@the_joshua_hill) came up with a new quant algorithm.

it consistently finds the optimal configuration:
a) in less time than SOTA
b) with more aggressive quant than SOTA
c) scoring higher on benchmarks than SOTA

achieving just one of the above is a feat on its own.

all three...excited for the paper to come out this week

[Jesse Genet @jessegenet]

Maybe the same unserious people playing these little business games shouldn’t also be trusted to tell congress what to do with AI, just a thought 🙃

[International Chess Federation @FIDE_chess]

#BeyondTheBoard 🇳🇴 Magnus supported the Norwegian football team at the FIFA World Cup, 🇮🇳 Pragg literally cooked, and 🇰🇿 Bibisara played the Kazakh national anthem on the piano (and that's after only 5–6 lessons!)

📹 players' social media

[abraham.py @def__abraham__]

Free Market Capitalism to the rescue 🦅 🦅

[Yuchen Jin @Yuchenj_UW]

Thank you, OpenAI, Meta, xAI, GLM, Kimi, Cursor, Cognition, for keeping the pressure on Claude to extend Fable access forever.

[Ethan Mollick @emollick]

Planning for how to use AI is a lot easier if there is some clarity about what to expect in the future.

Even a "we fully intend to keep extending this week by week but may need to stop under the following conditions and here is what the current status is" would be better.

[Perpetual Ogbiyoyo @guitargirrl]

We have our first winner! 🎉♟️

Thanks to @thegiftofchess , we’re giving away 10 chess sets to the first 10 people who show up and donate to Promoting Queens today.

One down, nine to go. If you’re near Washington Square Park right now, come play and claim yours.

Donate: https://t.co/eK9QJ67wzU

[Anthony Azekwoh @AnthonyAzekwoh]

The Grandmasters
Out now♟️
@Tunde_OD @chessinslums
https://t.co/LvaL7JzM42

[abraham.py @def__abraham__]

@amasad I have created Computer Vision models that can see and recognize chessboards, pieces and positions, feeds them into StockFish so you can play the chess engine over the board.

I’d love to see where models can explain the logic behind their moves in detail.

[abraham.py @def__abraham__]

“And for my next feature…”

You can now set your phone on a tripod and play StockFish Over The Board. The engine will call out the moves and correct you when you place a piece on the wrong square.

This would be available on my next update to KnightVision.

You can get it on the App Store:

https://t.co/ZbASoxIbFl

[Amjad Masad @amasad]

Fun to see Replit's computer use model play against my new chess engine

[abraham.py @def__abraham__]

This is the real benchmark on whether other models are good or catching up.

[Claude @claudeai]

We're extending Claude Fable 5 access on all paid plans, as well as keeping Claude Code’s weekly rate limits 50% higher, through July 19.

[Amjad Masad @amasad]

Vibe Research

Fine-tuning a Qwen-8b model to play chess on Replit. Running 3 parallel branches with different experiments and making real progress.

It's amazing how far models have come in their ability to do ML (they used to be really bad at it). So now someone with good intuition to guide the process could do interesting ML work, even if they have never done it before.