Ikbel retweeted
i recently saw e-means and just learned about Leiden communities and then you can put them together for benefits at 100m+ scale.
Leiden finds nicely shaped clusters in a neighbor graph, but is slow / memory constrained. e-means is super fast but cuts the space into even cells. We can calculate centroids with e-means and then partition those into communities. new data comes in can be assigned to a cell and hence a cluster super quick too
Ikbel retweeted
one of the most brilliant takes i've seen as of late on HCI in the era of LLMs
Attention steering by Sun Chuanqi. Actually operating an LLM's attention and token generation. the concept, UX and sound design are just amazing. There are still so many unexplored patterns. this one's particularly inspiring
github.com/chuanqisun/spaces…
Ikbel retweeted
an MIT researcher built a realtime LLM vector steering tool and it’s the coolest thing I’ve seen someone build in a very long time.
this is so brilliant. the primary use case I see this realistically used for is long form writing tasks. some might find it controversial, considering it plays into llm-generated writing. however in my opinion it kind of solves for some amount of the tension around the issue by making the process multiplayer in a new way. the human and the AI become collaborators, sort of.
idk this sh*t is so cool though, man.
(@ far__behind on instagram)
Ikbel retweeted
I updated my quantization compiler to use e-means as a replacement for k-means in codebook generation and got Qwen3.5-4B from 8.41 GB to 1.23 GB while retaining 91.9% of original performance which I think is approximately SOTA, at least comparable to bonsai 2 but I think we could go lower if someone wanted to donate some GPU time since this is just a hobby for me.
at this size, it's small enough that memory bandwidth isn't the bottleneck to decode speed so I'm interested in seeing how fast we can get larger models on consumer hardware. link below.
I have a feeling that the pool of ideas that AI models are proposing is shared.
In other terms, when you ask AI to do something in a specific way and share some ideas, in the next some months that same idea will be proposed to someone else.
Convergence on the way.
61
Reprogrammed my cursor to do two things that have been helpful:
- carry a screenshot (double-click to drop, shake to toss)
- jot a note in any margin (keeps the time and source)
The cursor is the one thing that's in every app, so it's extremely suitable for cross-app pains.
Ikbel retweeted
For anyone questioning whether this post is overhyped, I think they are actually *underselling* the approach.
We did similar experiments and found that Claude models can write entire, functional neural net weights from scratch! csinva.io/weight-space-progr…
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.…
Ikbel retweeted
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.…
Ikbel retweeted
So one of our ambassadors @divagr1925 has been working on something cool
He found a way to use Qwen's Gated DeltaNet state as memory, and built a 32-bit RISC-V CPU directly into the model's weights. So instead of relying on predicting the answer to a calculation, the model can execute instructions internally. It can also run C and MicroPython programs.
He calls it Qwenputer, love the name btw. The results so far are pretty impressive. On one arithmetic benchmark, accuracy jumped from 48.6% to 100%.
Pretty cool to see developers experimenting with Qwen's architecture. Great work, my friend!
效果最好的 PNG 压缩器已经开源啦!欢迎 star 和反馈 💗: github.com/heydeer/pngdeer
同质量压缩率比 TinyPNG 好 8%-12%,超过其他所有开源的压缩器。可以直接浏览器上测试: heydeer.ai/open-source/pngde…
Ikbel retweeted
If you've ever wondered why smart TVs are so cheap, this is why.
@log4j rooted his Samsung and found an always-on hot mic, automatic content recognition, geolocation, network mapping, ad targeting and user profiling, shared with Samsung's "400 partners".
A lot of it can be blocked in <15 min, wrote a basic guide below.
What your TV is doing
The main culprit is ACR (automatic content recognition). Your TV screenshots what's on screen, fingerprints it, and sends it home to build your ad profile. Samsung does it every 500ms, LG every 10ms [1].
What is insane: it happens on connected devices too.
Plug in your laptop, your console, your work machine, and your TV can fingerprint that via HDMI [1].
He also found that if your Samsung doesn't have a mic, it'll look for one from a connected soundbar or bluetooth device.
teapaw.org/log4j/status/211376224…
It's on by default, and you likely agreed to it on the setup screen you clicked through in 10 seconds.
Step 1: Turn off ACR
Samsung: Terms & Privacy → Viewing Information Services
LG: Live Plus
Vizio: Viewing Data
Roku: "Use info from TV inputs"
Sony/Google TV: Samba Interactive TV
Menus vary by model, but those are the names to look for. While you're in there, opt out of interest-based ads, reset your advertising ID, decline every optional privacy agreement, turn off voice recognition, and delete apps you never open.
Step 2: Give smart devices their own network
Put the TV, cameras, speakers and vacuum on your router's guest/IoT network. The logic is simple: these devices rarely get patched. If one gets compromised (getting drastically more frequent with AI), it shouldn't be able to see your other devices or home network.
Step 3: Block trackers for the whole house with DNS-level filtering
NextDNS (cloud-based) is easy and has a solid free plan. Pi-hole is a little bit more work if you prefer self hosting. For simplicity we'll focus on the former.
1. Sign up at nextdns.io (free tier is plenty)
2. Privacy tab: add a blocklist (OISD is a good bet) and turn on Native Tracking Protection for Samsung, LG, Roku, Amazon (see attached image)
3. Setup tab: copy your endpoint
4. Paste it into your router's DNS settings as DNS-over-HTTPS/TLS (or use the IPs + "Link IP" if your router only does plain DNS)
5. Bonus: block outbound port 53 except to your router. Some devices hardcode their own DNS to get around you
Then open the NextDNS logs and watch how often your TV phones home. It's eye-opening. If an app or update breaks, find the domain in the logs and allowlist it.
Why they do this
TV prices are down 97% since 2000 [2]. Most of that is better manufacturing and brutal competition. But once margins on the hardware hit zero, the cost moved to you.
In Q3 2024, Vizio lost $6.7M selling TVs and made $115.8M in gross profit from ads and viewing data [3]. They've said on earnings calls that they keep TV prices low to grow the audience they sell to advertisers. Walmart bought them in 2024.
This isn't a conspiracy theory either. Vizio paid $2.2M to the FTC in 2017 for tracking 11 million TVs without consent. Last December, Texas sued Samsung, LG, Sony, Hisense and TCL over ACR. Samsung and LG have since agreed to stop collecting that data without explicit consent [4].
TL;DR
1. ACR off
2. Ads + voice off
3. Smart devices on their own network
4. NextDNS filtering on the router
Sources:
[1] ACR study: arxiv.org/abs/2409.06203
[2] Cost of US consumer goods: aei.org/economics/chart-of-t…
[3] Vizio Q3 2024: sec.gov/Archives/edgar/data/…
[4] Texas v. Samsung: texasattorneygeneral.gov/new…
Ikbel retweeted
Introducing Pixel - a library for building GUIs in the terminal
- 120 FPS pixel-perfect graphics
- Embeddable inside existing TUIs
- Built in chromium webview component
pixelgui.com
Ikbel retweeted
introducing nanoui:
desktop apps in Zig, drawn on the GPU. Natively.
super low latency 0.06 ms frames
a SUPER nano 0.7 MB binaries
AND a 0% CPU when idle
WITH zero dependencies AND no webview
every frame of this video was drawn by nanoui itself(without a window).
Ikbel retweeted
D’un côté, on reste tourné vers le passé en ressuscitant de vieux modèles sans réelle valeur ajoutée et de l’autre on se lance dans les voitures volantes.
Reste à voir comment tout ça va se terminer…
Ikbel retweeted
Training Gaussian splats on a PS5 with Spirula Studio, running natively on the console’s AMD GPU via the new PS5 Vulkan port.
Impasse Saint Claude in Paris, captured a few years ago with @insta360
Ikbel retweeted
Introducing Graphical.
It's a simple, powerful, and fun tool to design visual languages, style components, and work with coding agents to create interfaces that look memorable and feel unique.
I hope you'll check it out at graphicalui.com!
Ikbel retweeted
Finally made this branching reply UI for real 🌿
Conversation trees can grow so wide!
but UI normally logs them down into cheap + obscurant linear scrolls
Growing some UI for spatially navigating the branching structure of reply threads
Interested in directly swiping through parallel piles of nested replies
(instead of the current "💬3" that can hide entire massive branches)
Made on-device, pushing @createwithplay to its limits
Ikbel retweeted
introducing Jevbox: an open-source document drive that organizes itself, answers questions, and cites everything (all powered by Jev).
TLDR;
- upload anything and Jev categorizes and files it into the folder tree
- search is hierarchical: Jev picks the folder, then the document, then the section. No embeddings, no vector DB
- every answer cites the page and section it came from
- permissions are enforced at retrieval. If someone loses access to a source, they lose access to answers built on it too
- ships with an MCP server, so your agents get the same permission-aware context
@andrewlu0 has been experimenting with Jev internally and wanted to see how far it could go on a real document library. It's built on Extend Parse + Jev, and fully open-source
The GitHub has instructions to self-host, or you can deploy it to Render in one click

































