HackerLangs
TopNewTrendsCommentsPastAskShowJobs

freakynit

1,538 karmajoined hace 11 años

Submissions

Show HN: A calculator CLI to help your agents perform calculations accurately

github.com
1 points·by freakynit·hace 4 días·0 comments

[untitled]

1 points·by freakynit·hace 4 días·0 comments

Show HN: Hacker-News Jobs

2 points·by freakynit·hace 6 días·1 comments

Show HN: HTTPS://HN-Jobs.com

hn-jobs.com
2 points·by freakynit·hace 8 días·0 comments

Is Trump-Netanyahu Rift on the Cards?

trump-netanyahu-rift.pagey.site
3 points·by freakynit·hace 20 días·1 comments

Renting vs. Buying – A case study for Bangalore

bangalore-property-buy-vs-rent.pagey.site
18 points·by freakynit·hace 22 días·19 comments

Show HN: JerrySniffs – MCP Server (and API's) for web and social media search

jerrysniffs.online
1 points·by freakynit·hace 24 días·0 comments

[untitled]

1 points·by freakynit·el mes pasado·0 comments

The Great Emu War (1932)

en.wikipedia.org
3 points·by freakynit·el mes pasado·3 comments

Nvidia – The Coming Short Squeeze?

linkshort.app
1 points·by freakynit·el mes pasado·2 comments

NSA Umbra UFO Files converted to Markdown and chat with them

1 points·by freakynit·hace 2 meses·0 comments

Show HN: Give This Markdown to Your Coding Agent Before Publishing to NPM

3 points·by freakynit·hace 2 meses·0 comments

Show HN: Nano-RAG – Agentic multi-hog retrieval without graph database

2 points·by freakynit·hace 2 meses·0 comments

Refactor: Unified Codebase for Better Performance

github.com
1 points·by freakynit·hace 2 meses·0 comments

All you need to know about the current hantavirus outbreak

sourceryintel.com
1 points·by freakynit·hace 2 meses·1 comments

The Chiplet Illusion: New Moore's Law or the Most Expensive Cover-Up?

sourceryintel.com
2 points·by freakynit·hace 2 meses·0 comments

Everything you need to know about recently released UFO files by Dow

sourceryintel.com
1 points·by freakynit·hace 2 meses·0 comments

Show HN: Chat with UFO Files

4 points·by freakynit·hace 2 meses·2 comments

The Instant Payments Race

nb1t.sh
2 points·by freakynit·hace 2 meses·0 comments

Show HN: Postlet – Tiny static blog generator with plugins and themes support

1 points·by freakynit·hace 2 meses·0 comments

comments

freakynit
·hace 24 horas·discuss
Cool... "vulture" looks pretty good... especially more now-a-days. Thanks..

I also maintain these sites for npm related attacks (since there have been so many since past 1.5 years):

1. Techniques exploited: https://npm-supply-chain-attack-techniques.pagey.site/

2. Documented attacks: https://npm-supply-chain-attacks-25-26.pagey.site/
freakynit
·anteayer·discuss
JetStream (not NATS) has a built-in key/value store. Relevant thread (old, from 2023): https://github.com/nats-io/nats.go/discussions/1507

someone did some benchmarks in 2025:

    nats bench kv put --size="128" --msgs 1000000 --storage file
    1m3s → Pub stats: 15,656 msgs/sec ~ 1.91 MB/sec
    
    nats bench kv get --size="128" --msgs 1000000
    59s → Sub stats: 16,720 msgs/sec ~ 2.04 MB/sec
I have personally not used this though.
freakynit
·anteayer·discuss
Excellent. We already tested at okayish scale of 5 million messages per minute, with each message being less than 1KiB. Do note that NATS can easily handle 100x this scale. It's just what we had tested during initial days.
freakynit
·anteayer·discuss
We are building something B2B expected to have good enough scale ... we had to choose between kafka, nats, redpanda and rabbitMQ .... We went ahead with NATS .... don't know why, but, when we went through all the docs, including setup and operational, NATS just felt right.
freakynit
·hace 3 días·discuss
Nop... when I opened the app, I found 100's of tasks within 100 meter radius.
freakynit
·hace 4 días·discuss
Just downloaded and made 15+ small contributions in the vicinity of my area. Very well built app. Super simple to use. And gamification is top-notch. Recommended.
freakynit
·hace 4 días·discuss
Automation script

    import time
    import pyautogui
    
    
    def rapid_click_for_300_seconds(interval: float = 0.01) -> None:
        pyautogui.FAILSAFE = True
    
        print("Clicking will start in 2 seconds...")
        time.sleep(2)
    
        print("Clicking started. Move the cursor wherever you want.")
        end_time = time.monotonic() + 300
    
        try:
            while time.monotonic() < end_time:
                pyautogui.click()
                #time.sleep(interval)
        except pyautogui.FailSafeException:
            print("Stopped because the cursor was moved to a screen corner.")
            return
    
        print("Finished clicking.")
    
    
    if __name__ == "__main__":
        rapid_click_for_300_seconds()
freakynit
·hace 5 días·discuss
I'm obviously biased as an India, but this is the best implementation. Very dense, yet, easier on the eyes.
freakynit
·hace 5 días·discuss
Palantir is giving me "Samaritan" vibes from the show: "Person of Interest".

For the ones who haven't watched this amazing show, here is a small Google AI summary:

Samaritan is the primary antagonist of the later seasons of the sci-fi series Person of Interest. It is a fictional, totalitarian artificial superintelligence created by Arthur Claypool. Unlike its counterpart, the Machine, Samaritan has no moral constraints, viewing human free will as a flaw requiring aggressive control and mass surveillance
freakynit
·hace 6 días·discuss
Bro wants a slave.
freakynit
·hace 6 días·discuss
Update: Added UMAP... this is so freaking cool.. thanks for suggesting.

It's on the insights page: https://hn-buddies.stupidlabs.lol/insights under "Author Map" heading.
freakynit
·hace 8 días·discuss
We may not see it yet, but, this is a way to capture government and use that to suppress competition and any dissent.

The moment the government owns equity, every regulatory decision becomes conflicted. The top 3-4 AI companies in whic the givernment owns equity, will become too politically entangled to fail or to let fail or be meaningfully competed against, or "allowed" to be competed against.

This will also lead to drasitc increase in surveillance state since givernment will now have a fudiciary, and "political" duty to increase dividends, "for citizens". One way to do that will be to purchase large amounts of surveillance equipment and tech disguised as "safety".

Any public dissent can now be indirectly suppressed by threatening with a perceived decrease in dividends or value of the sovereign wealth fund, thus, affecting healthcare, and other public expenditure.

This also sets a very bad precedent for other governments to follow. This will end up with givernments of major nations to act in similar fashion, thus, expanding this dystopian setup further.
freakynit
·hace 9 días·discuss
Might be weird, but, I have heard this "2012" reference many many times when things started to change (for worse), and then "2020" when it all started to go downhill with an ever-accelerating pace. Very consistently these two years from multiple people.

Any rational explanation for this?
freakynit
·hace 12 días·discuss
Meme context: https://x.com/BrunoCendon/status/1926883516890898637
freakynit
·hace 12 días·discuss
80B + INT4 + speculative (FP8 mode) => 72,188 tokens/s effective

..da fck!!
freakynit
·hace 12 días·discuss
Same.. although rare, but have observed twice till date.

Some blog post I read few weeks back said that DSV4Flash in xHigh effort beats even the pro model in xHigh effort.
freakynit
·hace 13 días·discuss
Recently read about some upcoming GPU chip vendor that speeds up multiply-and-add (the core of [pretty much any neural network, including LLM's) oepration by transforming them to log scale where the multiplication changes to addition.

There issue was the size of such log tables that would be needed, and hence, they settled for some trade-off.

My question is: can the expensive log-domain addition / correction function be implemented as fixed hardware lookup tables or approximate units?

Update: found the vendor: https://www.tensordyne.ai/silicon-and-math
freakynit
·hace 13 días·discuss
Just tried. Bare minimal and does exactly what's needed and nothing else. Love it. Thanks..
freakynit
·hace 14 días·discuss
You could produce a completely fcked up model using AI, and also a completely awesome one using AI. So, the "overstating capabilities" part really have nothing to do with this discussion.

As for the claim part around AI being used to assist further AI development, I have already told to read Anthropic's blog post. You can take help from chatgpt, if needed.
freakynit
·hace 14 días·discuss
You do realize this statement about "overstating model capabilities" has nothing to do with AI assisted AI development, right?