Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
auspiv
searching Neon…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
1.
▲
by
auspiv
8d ago
This team is in a tough spot - they'll need to innovate and get their models in a good spot quickly. They are racing the clock against various similarly sized models being burned into ASICs and running at a similar speed (or faster).
2.
▲
by
auspiv
20d ago
I think the question is probably more along the lines of - why not do a database with 100 TB of storage/records instead of a cache? tomato / tomato.. especially with smart caching in front of database. 100TB of flash is a good bit
3.
▲
by
auspiv
1mo ago
I'd take qwen3.6 (3.8 as of tomorrow) 27B running at 17k per second first on the way to Sol/Fable! And then dsv4-flash-0731!
4.
▲
by
auspiv
1mo ago
Opus 4.5 level of performance is also accessible with deepseek-v4-flash-0731 (0731 being the july 31 update) which is much, much, much smaller. 2x RTX pro 6000 blackwell can run it. 4x can run it very comfortably
5.
▲
by
auspiv
1mo ago
https://www.nhc.noaa.gov/
6.
▲
by
auspiv
1mo ago
Modern dell desktops (like 8th gen+) idle around 10W. Raspberry pi isn’t saving much
7.
▲
by
auspiv
1mo ago
“Search session history for when we discussed XYZ” https://github.com/nerdyaustin/memory_mcp You can skip all the sync stuff, not necessary at all
8.
▲
by
auspiv
1mo ago
which would make it the 37th+ spacecraft to hit the moon (most S-IVB rocket stages and Lunar Module ascent stages impacted the Moon - not tracked individually on that page) https://en.wikipedia.org/wiki/Category:Spacecr
9.
▲
by
auspiv
2mo ago
Well when a huge part of potential revenue is all in on Bedrock... you need the harness to be able to talk to Bedrock. And Vertex. And all the other places these models are hosted. And allow for proxy because many businesses do not all dire
10.
▲
I sent Claude Opus 5 '–-' and it wrote me 5k tokens about a cartographer
(austinsnerdythings.com)
16 points
by
auspiv
2mo ago
|
6 comments
11.
▲
by
auspiv
2mo ago
This is/was classic market segmentation. Want ECC? Pay up for Xeon (server/workstation CPU)
12.
▲
by
auspiv
2mo ago
Statistically speaking if each part of the Luna run has a 90% chance of being correct, 5 of those is 0.9^5 = 0.59 = 59%. Or one Sol run being maybe 95% correct? Exact numbers vary of course. But then again having sol verify at end may be ch
13.
▲
Show HN: DIMMsum – price tracker and sold-price history for used server RAM
(dimmsum.com)
1 points
by
auspiv
2mo ago
|
0 comments
14.
▲
by
auspiv
2mo ago
you don't. they are distinct entities. https://platform.claude.com -> api keys
15.
▲
by
auspiv
3mo ago
The current natural gas price in West Texas (the WaHa hub, north of Coyanosa, TX) is negative. And has been for a while. The price peaked (dipped?) to -$9/MCF a couple months ago. That means gas producers had to pay $9 per MCF for it t
16.
▲
by
auspiv
4mo ago
Wow, not one mention of the env vars that have a far greater influence on how the models actually work under the hood - https://code.claude.com/docs/en/env-vars Very important for bedrock deployments and other not
17.
▲
by
auspiv
4mo ago
GPT-5.5 is a solid leap with Codex or other harnesses. Opus 4.7 I still don't understand how people use... I tried it for a day or two, have tried it for a few hours every week or so since release, and still use 4.6 as daily driver (wi
18.
▲
by
auspiv
4mo ago
And their spreadsheets don't?
19.
▲
I swept all 667 layer duplication configs for Qwen3-4B on 1x RTX 3090
(austinsnerdythings.com)
3 points
by
auspiv
5mo ago
|
0 comments
20.
▲
One Layer, +12%: What 667 Configs Reveal About Small LLM Anatomy
(austinsnerdythings.com)
2 points
by
auspiv
5mo ago
|
0 comments
21.
▲
by
auspiv
5mo ago
probably same thing for boeing 787 jets - https://www.theregister.com/2020/04/02/boeing_787_power_cycl... says 51 days, which would be an interesting number of (milli)seconds
22.
▲
by
auspiv
6mo ago
apparently you can straight up duplicate/add/rearrange layers without changing any of the weights and get better results as well - https://dnhkng.github.io/posts/rys/
23.
▲
Show HN: Trained YOLOX from scratch to avoid Ultralytics (aircraft detection)
(austinsnerdythings.com)
2 points
by
auspiv
7mo ago
|
1 comments
24.
▲
Show HN: Trained YOLOX from scratch to avoid Ultralytics (iOS aircraft detect)
(austinsnerdythings.com)
3 points
by
auspiv
7mo ago
|
0 comments
25.
▲
by
auspiv
8mo ago
There have been a couple "studies" and comparing various frontier-tier AIs that have led to the conclusion that Chinese models are somewhere around 7-9 months behind US models. Other comment says that Opus will be at 5.2 by the ti
26.
▲
by
auspiv
8mo ago
Wow, 249 CHF for 8x fans is insane. The grip Noctua has on people! Nice workstation.
27.
▲
by
auspiv
8mo ago
Using this with tmux and various VPN tech. Main issue is scrolling. Termius + tmux don't scroll very well. And I've been led to believe tmux is necessary to keep sessions open when I turn off my phone screen
28.
▲
by
auspiv
9mo ago
Thanks! How I eventually found it was stripping stuff back layer by layer. And by that I mean I started with the raw camera feed and got to where things worked well in a different swift view. And then from there, peeled stuff back from the
29.
▲
Show HN: I vibe-coded an aircraft AR tracking app and wasted weeks on AI bugs
3 points
by
auspiv
9mo ago
|
2 comments
30.
▲
by
auspiv
9mo ago
ZScaler (at work) classifies this whole site as Website blocked Not allowed to browse Weapons/Bombs category You tried to visit: https://relaxing.run/blag/posts/top-gun-landing/
More ›