witnessme

5 ms·

1.
by witnessme 4mo ago
Sure it is small, 3B. But on Pi Zero, I can tell you from my experience, you'll be disappointed.
2.
by witnessme 6mo ago
Interesting. How does this foundational model compares with other LLMs?
3.
by witnessme 7mo ago
This is a much needed idea. I'd rather enforce this rule in my community even if it means less comments. The only concern is that I might miss out on some opinions that were horrible but still better than nothing.
4.
by witnessme 8mo ago
Exactly. Most folks are trying force openclaw on their workflow which would have been way cheaper and better without it
5.
by witnessme 8mo ago
The concern is real but the local solution is not ready. The author does not seem to think about that from the perspective of an "average consumer". I have been running my personal AI assistant on a consumer-grade computer, for al
6.
by witnessme 8mo ago
Agree with the points. Cybersec audits are mandatory for insurance companies in most countries. This list need to be expanded.
7.
The death of the two-week sprint(blog.invidelabs.com)
1 points by witnessme 8mo ago|0 comments
8.
Ask HN: My post does not show up in /newest again
1 points by witnessme 8mo ago|2 comments
9.
by witnessme 8mo ago
That would be 8 years after math + humor peaked in an article about singularity
10.
Substack confirms data breach affects users’ email addresses and phone numbers(techcrunch.com)
100 points by witnessme 8mo ago|44 comments
11.
by witnessme 8mo ago
I am still confused for days whether this is a real news or a hoax. Only a substack user saying they received this email. I did not. And there is no official statement by Substack. What is really going on here?
12.
by witnessme 9mo ago
Can't agree more
13.
by witnessme 9mo ago
Cracked at "I need to contribute my own beating so this horse is really dead"
14.
From fear to pain: NPM trusted publishing and oidc(github.com)
1 points by witnessme 9mo ago|1 comments
15.
by witnessme 9mo ago
2025 was the year when we saw extreme impact of supply chain attack. npm did something last month to counter this by retiring old auth tokens in favor of a new authentication method they call trusted publishing, it uses OIDC to authenticate
16.
by witnessme 9mo ago
Going to try this one soon. Great work OP.
17.
by witnessme 10mo ago
This is a novel idea. Somewhere between the extremes of being useful vs being an overkill. More towards overkill because of its dependency on a new app/browser that needs to be installed. But I'm looking forward to more developmen
18.
by witnessme 11mo ago
Noted. I can see it on the link you shared. Earlier, I did check until the last page (90+ items) and did not find it. I then checked couple of more times thinking I made mistake. I see that you're using /shownew, I used /show
19.
Ask HN: My post not shown in showhn
1 points by witnessme 11mo ago|2 comments
20.
Show HN: Resilient LLM(github.com)
1 points by witnessme 11mo ago|0 comments
21.
by witnessme 11mo ago
The same feeling non-native English speakers have battling native English speaker bias
22.
by witnessme 1y ago
A simple word2vec embedding with continuous bag of words (CBOW) training is enough and beats all other complex solutions at rhe performance as well as cost Reference: https://blog.invidelabs.com/how-invide-analyzes-deep-work
23.
by witnessme 1y ago
Claude team has been killing it with the new impressive releases since last week. And this one looks most promising.
24.
by witnessme 1y ago
Why is featured as #1 on the frontpage. I get it, nice piece of satire and a bit controversial. But it is not productive at all.
25.
Tell HN: Online coding support we used to know, is dead
2 points by witnessme 1y ago|2 comments
26.
by witnessme 1y ago
Does Decode project overlaps with Solid project (Sir Tim Berners Lee's project with similar objective)? P.S. Thanks for sharing Decode, it looks good, I will be diving deeper into it this weekend. TIA.
27.
Android for PC Is Coming(theverge.com)
3 points by witnessme 1y ago|0 comments
28.
by witnessme 1y ago
For folks wondering whether to read or not, here is the conclusion from the paper verbatim > We have demonstrated that a single transformer-based model can effectively learn and predict the dynamics of diverse physical systems without ex
29.
by witnessme 1y ago
Oh CERN folks, you beautiful people
30.
by witnessme 1y ago
Don't try 12-20B on 16GB. You should stick with 4-8B instead. You'll get way too slow tps and marginal perf improvements going higher on a 16GB machine.

More ›