bratao

I started creating internal models using it, then the Gated Deltanet 2 came out( https://arxiv.org/abs/2605.22791), and it seems like an evolution of it in expressiveness. And in our tests it is really better than.

I know this is a common trope here, and yeah, plenty of MBA-type CEOs are trash. But speaking as a founder/CEO of a small company, the quality gap between an average hire and top talent at the director level is night and day.

Proven talent comes with a massive price tag, but it's almost always worth it. The times I regretted going with the 2nd or 3rd place candidate because they were cheaper ended up setting us back by months.

The same logic applies to CEOs. Moving the needle for a massive company by even a few percent is worth billions. Just look at Lisa Su or Steve Jobs.

This could be controversial, but when hiring and I see anyone that has multiple parallel jobs, it is an instant reject. Maybe this could be it? This sentiment is shared by many other persons I talked, and come from bad experience in the past.

Not affiliated but I been using https://github.com/superdoc-dev/superdoc and it is very good and compatible with many docx features.

I´m super curious about those "Two Weird Tricks". I would like that you would release more. It remember me the MiniMax Sparse Attention https://arxiv.org/html/2606.13392v1

It is super strange that all last (3?) releases they keep comparing older models such as Opus-4.6.

Look like the Founder and CTO account has been compromised. https://github.com/krrishdholakia

One interesting point of GitLab for me is the self-hosted version, including the AI features (Duo) that can also be self-hosted and you can bring your own OpenAI/Anthropic key.

From my experience MariaDB is not necessarily better than MySQL. The 8.x line brought many interesting features. I dream on switch to Postgres, and try every year but for my use case MySQL is still superior (100Bi+ rows for large texts, and heavy modified - I´m also space constrained - So I need data compression and the VACUUM are not good.)

The percona distribution is very good!

The danger of Internet centralization in Cloudflare

One thing I don't understand about Nvidia’s valuation is that right now a small number of algorithms have 'won,' such as Transformers. The data is very important. Compared to the past where customized code was much more common, such as modeling code and HPC, the ecosystem was very important and it was almost impossible to implement all CUDA and related code.

Competitors now only need to optimize for a narrow set of algorithms. If a vendor can run vLLM and Transformers efficiently, a massive market becomes available. Consequently, companies like AMD or Huawei should be able to catch up easily. What, then, is Nvidia’s moat? Is InfiniBand enough?"

The link is off. This link works https://api-docs.deepseek.com/updates#deepseek-v31-terminus

I use Rancher for a hosted Kubernetes cluster on top of dozens of dedicated servers, and so far, it has been super nice. What are the alternatives for CI/CD for a small team (30)?

I'm really curious about the history of spaCy. From my PoV: it grew a lot during the pandemic era, hiring a lot of employees. I remember something about raising money for the first time. It was very competitive in NLP tasks. Now it seems that it has scaled back considerably, with a dramatic reduction in employees and a total slowdown of the project. The v4 version looks postponed. It isn't competitive in many tasks anymore (for tasks such as NER, I get better results by fine-tuning a BERT model), and the transformer integration is confusing.

I feel sad and disappointed in Microsoft for letting the entire Faster CPython team go. I was a big supporter, always leaving positive comments and sharing news about their work. I'd figure the team paid for itself in goodwill alone. What a letdown, Microsoft. You ought to do better.

You are even luck to be able to verify. Mine give me an error about "Session expired" for months!! Support do not reply.

If the numbers are true, this would be one of the more successful attacks in history. Drones are changing the whole dynamic of wars.

This checks if the Library builds or if it is really compatible/works with multi-threading?

This is a common mistake and very badly communicated. The GIL do not make the Python code thread-safe. It only protect the internal CPython state. Multi-threaded Python code is not thread-safe today.

If this tariff situation really impact the US economy, when should we start to see layoffs?

For all LLMs, I´m using a simple prompt with the complete code in triple quotes and the command at the end, asking to output the complete code of changed functions. Then I use Winmerge to compare the changes and apply. I feel more confident doing this than using Cursor.

From my use case, the Gemini 2.5 is terrible. I have a complex Cython code in a single file (1500 lines) for a Sequence Labeling. Claude and o3 are very good in improving this code and following the commands. The Gemini always try to do unrelated changes. For example, I asked, separately, for small changes such as remove this unused function, or cache the arrays indexes. Every time it completely refactored the code and was obsessed with removing the gil. The output code is always broken, because removing the gil is not easy.

Am I in a bubble? There is any justification about Trump actions? From my perceptive(Brazilian) the US is dissolving all the soft-power it had in a record-breaking speed.

I´m super excited, sleepless for a couple of days already. I´m trying to use all tricks possible to improve a Sequence Labeling using Conditional Random Fields. I need to NER billion of documents, and need to be fast. CRFSuite is a workhorse, and a baseline very hard to beat with speed and precision. But with o3 I´m created a frank-stain with many tricks such as CRF with variable order, feature interactions, bidirectional, jointly learning with word embeddings. The precision is already over than CRFSuite. And I believe that would be better than many other solutions such as bi-lstm-crf. Definitely much faster.

Now i´m trying to port to Cython to make as fast as possible. Here o3 is almost useless, but I´m progressing.

Every six months, I explore switching from MySQL to something new for a more modern tech stack. However, MyRocks (https://docs.percona.com/percona-server/8.4/myrocks-index.ht...) is truly impressive. It allows me to efficiently compress my text-rich rows.

What would be the most performant way to run a inference using RWKV? Do you have and speed comparison to a similar sized transformer?

I have a task(OCR cleaning) that I´m evaluating faster options and look like RWKV would be a nice alternative.

Yeah, super strange. One cannot finish a sentence without the other interjecting.

I hate PHP as a language. However as an Infra guy, it is very performant and easy to scale. Laravel is also an incredible framework.

Unfortunately it is not Opensource yet - https://github.com/fireducks-dev/fireducks/issues/22

SPANN is also implemented in the open-source Vespa.ai