Hacker News Reader

You mean the one they copied from Microsofts paper? (properly cited as well)

cpldcpuon Show HN: Mini-AGI – Dynamic continual learning model trained on 8GB VRAM

Is this architecture actually able to generalize or is it mostly based on memorization? Have you tried some basic tasks that require generalization? e.g. number addition etc?

cpldcpuon LLMs are real, AI is fake

Well, that is what video models do.

But keep in mind: So far there is no way to train the models while completely avoiding memorization and only including generalization. That would be a great way to avoid any copyright issues, but all attempts I have seen so far were fairly limited.

cpldcpuon LLMs are real, AI is fake

>Notably, the neural network component behaves like a compressed index of the training data.

The "Compression" part is indeed about the model finding order in the training data. But this is not about applying some predefined compression algorithm on it, but by actually learning an algorithmic representation of the data.

This has nothing to do with "consulting the training data", because the training data cannot be reconstructed based on this information.

Roughly speaking, there are two types of information being stored in the NN: Memorization and generalization, or, shannon entropy and kolmogorov complexity.

This paper is highly underrated: https://arxiv.org/pdf/2505.24832

cpldcpuon Flux 3

its doing video, audio, images and motion. I think that counts as multimodal.

cpldcpuon Materials innovation has a scale-up problem, not discovery

>When Intel finally shipped it at the 45-nanometer node in 2007, Gordon Moore called it the biggest change in transistor technology since the late 1960s. The breakthrough was not the material. It was learning how to process the material at scale.

Its curious that they picked this example. The challenge with HKMG was not the material itself, but how to integrate into into the transistor stack.

There were two completely different approaches: Gate first and replacement gate. Gate first is what the industry was already using for silicon oxide so everybody tried to go with as little change as possible. Only intel decided for replacement gate, which worked much better and reaped some other benefits on the way.

This was a watershed moment in the industry and ultimately led to some of the players dropping out of the cmos race.

But is this really a "scale-up" problem? It required development of novel manufacturing processes (atomic layer deposition), but was still mainly a process integration and device engineering topic.

The part of the thesis I have to agree with is that there is a data problem. The development above relies on executing lots of time consuming and tedious split experiments that often cannot be parallelized. The outcome of this relies heavily on the experience and diligence of the experimenters.

It's probably well suited for an "autoresearch" approach, bridging to the phyiscal world and dealing with the timescale is the challenge.

cpldcpuon OpenAI unveils its first custom chip, built by Broadcom

hm.. has been quite a while for me. The good thing about the Tang Nano is that it is supported by the Yosys open source toolchain. There are quite a few resources on the web when you search for the combination.

cpldcpuon OpenAI unveils its first custom chip, built by Broadcom

I had Opus 4.5 design an LLM inference engine in verilog, including firmware and automated verification a while ago: https://github.com/cpldcpu/smollm.c

It's of course far from optical. But lowering the implementation through the abstraction levels turned out to be extremely powerful.

cpldcpuon The 29th International Obfuscated C Code Contest (IOCCC) 2025 Winners

CISC view: Its another adressing mode.

RISC view: SUBLEQ is already four instructions (2x memory access, alu, branch)

cpldcpuon The Russian who invented semiconductors 25 years before the USA

Great article!

Yeah, that pattern can be seen everywhere in semiconductors. E.g. the transistor invention vs. Lilienfeld, Heil, Matare etc. So the scope is more narrow than "Inventend Semiconductors".

Generally, there seems to be a tendency to disregard discoveries from outside the US. I think this pattern can still be observed today...

Other examples: Invention of light bulb, telephone.

cpldcpuon The 29th International Obfuscated C Code Contest (IOCCC) 2025 Winners

Wow! And it also implements a very interesting variant of SUBLEQ that is turing complete.

>This VM implements an OISC - a One Instruction Set Computer. That instruction takes three signed 32-bit operands, a, b and c, and runs a program from memory m[] as follows:

1 PC (program counter) starts at 0

2 Fetch the next instruction (32-bit signed operands a, b and c)

3 If the low bit on any operand is set, remove it, and replace that operand with m[operand] i.e., a dereference of that address

4 Set m[b] = m[b] - m[a]

5 If m[b] is 0 or negative, set the PC to c, otherwise increment PC by 3 words

6 Go to step 2

cpldcpuon The Russian who invented semiconductors 25 years before the USA

The early discovery of light emission from silicon carbide long before the first LEDs is a very interesting finding, worth pointing out.

But alas, as ever so often, the article turns this into a hyperbole. The premise from the title does not check out at all.

>The Russian who invented semiconductors 25 years before the USA

https://en.wikipedia.org/wiki/Semiconductor#Early_history_of...

cpldcpuon I am against GenAI and everything it stands for

What do you mean with "open-source"? Of course, the inference code for all the open weight models is publically available - see llama.cpp or hf transformers.

There are, however, very few models where also the full training pipeline is available. Olmo by AI2 comes to mind.

cpldcpuon Hosting a website on an 8-bit microcontroller

Wow, nice! It's an honor.

I guess nowadays one could use some of the 32bit WLCSCP microcontrollers to easily beat this.

cpldcpuon Hosting a website on an 8-bit microcontroller

More than 25 years ago, there was a show off, of building the smallest web server:

https://web.archive.org/web/20000815063022/http://www-ccs.cs...

Someone with an ACE1101 microcontroller "won". I can't find the original articles, but there is also this:

https://conceptlab.com/fly/

Webserver on a fly...

cpldcpuon Building the TD4 4-Bit CPU

From the blog post: https://github.com/wuxx/TD4-4BIT-CPU

cpldcpuon Building the TD4 4-Bit CPU

Nice!

cpldcpuon Building the TD4 4-Bit CPU

But why is this always the first comment on custom CPU builds? Can't there also be other designs out there?

cpldcpuon Eden AI – European Alternative to OpenRouter

nah... their founders and most of their employees are in france.

Very ill-suited comparison to IBM.

cpldcpuon Eden AI – European Alternative to OpenRouter

https://huggingface.co/docs/inference-providers/pricing

It's well buried though. Does not seem to be a focus of theirs.

cpldcpuon Eden AI – European Alternative to OpenRouter

Great to see this!

Worth mentioning that Huggingface already offers a similar service. And they are also European:

https://huggingface.co/docs/inference-providers/index

https://huggingface.co/inference/models

cpldcpuon Martin Galway's music source files from 1980's Commodore 64 games

I had to give it a try.

Claude, the ole cheater, recognized what the file was, downloaded the psid from the web, found a wasm sid player and built a website around it:

https://claude.ai/public/artifacts/df6cdcae-08dc-452b-ba19-f...

https://claude.ai/share/4dd36c16-bc62-445a-b423-ad4637f06432

GPT-5.5 built a lot of python scripts to extract the music data. Strudel implementation failed, but I then asked it to build a website:

https://ubiquitous-vacherin-8e7993.netlify.app/

This is a translation of the music into javascript based on the assembler source.

Really impressive on both accounts. Some iterations were requied for both.

cpldcpuon Exploiting the most prominent AI agent benchmarks

Yes, marks of AI all over the place. Also the SVGs.

>No solution written, 100% score.

Its weird. Turns out that hardest problem for LLMs to really tackle is long-form text.

cpldcpuon Show HN: I built a tiny LLM to demystify how language models work

Love it! Great idea for the dataset.

cpldcpuon Building an FPGA 3dfx Voodoo with Modern RTL Tools

cpldcpuon How Taalas “prints” LLM onto a chip?

They mentioned that they using strong quantization (iirc 3bit) and that the model was degradeted from that. Also, they don't have to use transistors to store the bits.

cpldcpuon How Taalas “prints” LLM onto a chip?

I wonder how well this works with MoE architectures?

For dense LLMs, like llama-3.1-8B, you profit a lot from having all the weights available close to the actual multiply-accumulate hardware.

With MoE, it is rather like a memory lookup. Instead of a 1:1 pairing of MACs to stored weights, you suddenly are forced to have a large memory block next to a small MAC block. And once this mismatch becomes large enough, there is a huge gain by using a highly optimized memory process for the memory instead of mask ROM.

At that point we are back to a chiplet approach...

cpldcpuon How Taalas “prints” LLM onto a chip?

It could simply be bit serial. With 4 bit weights you only need four serial addition steps, which is not an issue if the weight are stored nearby in a rom.

cpldcpuon Reproducing DeepSeek's MHC: When Residual Connections Explode

Thanks! Would be quite interesting to see how this fares compared to mHC.

I noted that LAuReL is cited in the mHC paper, but they refer to it as "expanding the width of the residual stream", which is rather odd.

cpldcpuon Reproducing DeepSeek's MHC: When Residual Connections Explode

May be worth pointing out, that this is not the first residual connection innovation to be in production.

Gemma 3n is also using a low-rank projection of the residual stream called LAuReL. Google did not publicize this too much, I noted it when poking around in the model file.

https://arxiv.org/pdf/2411.07501v3

https://old.reddit.com/r/LocalLLaMA/comments/1kuy45r/gemma_3...

Seems to be what they call LAuReL-LR in the paper, with D=2048 and R=64