netdur

I find helpful ads on Google Search sometimes, and it can be the easiest way to get results, but most of the time, ads (and SEO) ruin search accuracy to the point that it's becoming totally useless

I am working https://vibu.app which is free digital voucher

and for fun, I am building yet another programming language!

Mistakenly, i thought it was about Rotten Tomatoes, and i started thinking about how a movie like Michael ranked badly, the critics missed the whole point of watching a movie, to be entertained, sadly, here on HN, sometimes we miss the point too, if that involves some names

Then do not use Android, some Chinese phone or iPhone

I am getting 21 t/s on Fold 7, 21 x 1.8 = 37.8 t/s compared to M1 Max's 54 t/s, that is impressive

why would the bbc publish an article about a man who believed a chatbot's story about someone coming to kill him?

hmm... at Q4_K_M, stock-style quantization is retaining ~99–99.8% of BF16 accuracy, AutoRound pushes that to ~99.4–100.n% (??) the gap is roughly 0.1–0.7 percentage points

https://github.com/intel/auto-round/blob/main/docs/gguf_alg_...

In court vs openai, Musk said Grok is partly trained on openai models, so it should be somehow similar to Chinese models in terms of performance and cost!

they should just swap it with Qwen 3.6 27B, no one would tell the different

Geopolic: A US-aligned Gulf state walking away from a Saudi/Russia-led bloc in the middle of a war, after deciding the bloc didn’t really have its back

Economic: it weakens OPEC’s pricing power in a way you might not see right away if Hormuz is closed, but it could really change the supply picture once things reopen

By asking Meta polity

there are two sides, such as how photos can stress citizens and act as propaganda, making them harmful to state interests, ultimately it is their country and their rules, not yours, regardless of how much you disagree with it

you are also missing the elephant in the room, whatsapp's claim of end-to-end encryption is a lie

didn’t we have several antitrust cases where a vendor used its monopoly to disadvantage rivals? did not anthropic block openclaw?

I am working on hugind, I have two goals:

- make it reliable to run LLM inference on company hardware, even when it is poor or outdated

- bring chaotic agentic behavior under control in business contexts

https://github.com/netdur/hugind

I tried TQ for vector search and my findings is not good, it is not worth it if you cannot use GPU, however I got same quality of search as 32f using 8bit quant

I wrote ann ext for sqlite, using tq, I do save a lot on space but 32f is still faster despite everything I have tried

code here https://github.com/netdur/munind/tree/main/src/tq

I wrote an ai agent that do chrome testing, yes, chrome MCP do work https://github.com/netdur/hugind/tree/main/agent/chrome_test...

Did it complain about copyright issues?

Gemini 3 long context is not good as Gemini 2.5

Now you know why Superbowl ads cost millions and bloggers are just bloggers

I have tried to solve the agent running wild, and I found two solutions, the first is to mount the workspace folder using WASM to scope any potential damage, the second is running rquickjs with all APIs and module imports disabled, requiring the agent to call a host function that checks permissions before accessing any files

--- [0] https://github.com/netdur/hugind

While there is nothing wrong about Saudi money, they are spreading FUD about open weights and wants to regulate and control how the LLM is used

i would love if you took the time to instruct claude to re-implement inference in c/c++, and put an mit license on it, it would be huge, but only if it actually works

calculators doing a perfect job did not end accountants' jobs, it made it faster.

they have been doomed for a while, it is just a matter of time, but honestly i like them better than the claude provider, if they can make openai profitable, that would be good for all of us, we don't want a world where gemini is the only winner or the chinese take over

Anthropic thinks highly of its "moat", yet it is spreading FUD to kill open weights

Ai studio with my custom prompting is much better than Gemini app and opus

i think the concern about software shifting toward ai design ignores that the web hasn't been human-first for a long time. most traffic is already machine to machine, like crawlers and ci pipelines. we’ve tolerated systems that are barely legible for years. anyone who has grepped through android studio logs knows that human readability is usually a tertiary goal at best. ai interacting with complex systems is just an evolution of the glue code we’ve always written.

as for who made it, utility usually matters more than where it came from. i used an agent for an oss changelog recently and it picked up things i’d forgotten while structuring the narrative better than i could. the intent and code were mine, but the ai acted as a high fidelity compressor. the risk isn't ai being everywhere. it’s the atrophy of judgment where we stop using it to support decisions and start using it to outsource thinking.

i don't understand americans, two years ago i wanted a tesla, now i want a byd, you've let down the only american company competing against the chinese, all because of trump and politics

I believe it is the system instructions that make the difference for Gemini, as I use Gemini on AI Studio with my system prompts to get it to do what I need it to do, which is not possible with gemini.google.com's gems

for document layout! did you have success understanding document layout using SAM