a11r

4 ms·

1.
by a11r 9d ago
Pretty impressive to see the amount of performance they can squeeze out of the same hardware. I suspect the same process will play out for all combinations of LLMs, inference providers and hardware stacks. This should bring down the cost of
2.
by a11r 9d ago
Back in 2008 Fujitsu has one of the best performing 10Gbps Switches. We were building 40 Gbps packet sniffers at Google (4x 10 Gbps NICs) and needed switches that could do things like mirror traffic across ports at line rate. Fujitsu was wa
3.
by a11r 23d ago
It is great to see another player introduce a fully open stack. Nvidia's Nemotron is the only other prominent one I know of. All that said, the headline claims do not match the self-reported performance. For example, the dense 32B mode
4.
by a11r 24d ago
Looks like the strategy of regular updates with incremental improvements is working out well. Interestingly, the biggest jump in Artificial Analysis Intelligence Index score is for reasoning level Medium ( 3.7 was 51, 53, 57 for Low, Medium
5.
by a11r 1mo ago
Quoting RGFusion from Reddit: LLMs run into an issue where the further you train a model, the more it overwrites facts with generalized concepts. You need the model to be able to do both. Intelligence arises from generalization, but without
6.
by a11r 1mo ago
There is a difference between needing frontier capability because one is solving a truly open ended problem, and needing a reliable workhorse model to do something well understood. Local models (like Qwen 3.8 27B) have gotten so good that t
7.
by a11r 1mo ago
Even a 4-bit quant of Qwen3.8 27b is indistinguishable from Gemini 3.7 flash in our internal tests. With an RTX5090 card and ninfer, you can get ~800 TPS token generation (c=8) and ~140 Tokens per second single stream.
8.
by a11r 1mo ago
M3 Pro 36GB. I am getting 17 tps with MTPLX.
9.
by a11r 1mo ago
Amit Patel was employee #7 at Google and built many things including the little geo display in the lobby that would show live search queries as dots color coded by language.
10.
by a11r 2mo ago
The pricing is 40% of Kimi K3 at the same Model size. It will be interesting to see benchmarks tomorrow and to see how the pricing evolves across this size class.
11.
by a11r 2mo ago
Seminanlysis is estimating sub $1 cost per MT for ~2Trillion models. The numbers change based on throughput and quant, but it is conceivable that provider costs at scale are low enough that even $2.42 per MT on GLM 5.2 (current best price)
12.
by a11r 7mo ago
This resonates with my experience. At Morph we use gemini for well specified point coding tasks, and it does very well across millions of lines of code every day. We also use claude code as an engineering tool for our own codebase and it do
13.
by a11r 7mo ago
Understood. I am looking for a side-by-side comparison focused on your feature set, not Github's. You answered it partially by calling out your focus areas. Github API reliability has been iffy for us, so it would be good to quantify t
14.
by a11r 7mo ago
It would be nice to see a side-by-side comparison with Github on pricing and features. We are using github and creating hundreds of repos everyday without any issues (except for the occassional API outages that Github has). Curious to see y
15.
by a11r 8mo ago
This is very nicely done. We have seen the same issue at a higher level of getting separators right when generating multiple files in a single inference call.
16.
by a11r 1y ago
https://house.fandom.com/wiki/Clueless
17.
by a11r 1y ago
Nice ! Curious what your prompt was.
18.
by a11r 1y ago
These are large coal seams with significant exposure to the atmosphere. See https://en.wikipedia.org/wiki/Jharia_coalfield for an example. That excavator in the picture is not trying to put out the fire, it is just min
19.
by a11r 1y ago
The system prompt in this experiment limits the solution to always spell out the concrete moves verbally. A human solving the Tower of Hanoi gives up around N=4 and goes off to invent a recursive solution instead. Prompted differently, the
20.
by a11r 3y ago
Gary Taubes has written multiple scholarly books on the subject. He has won the Science in Society Journalism Award of the National Association of Science Writers three times. I think a training in one scientific field does qualify a person
21.
by a11r 3y ago
For people with mobility challenges, or those with communication challenges, being able to look at a map of the airport and plan in advance is useful.
22.
by a11r 4y ago
"Yunnani" medicine comes from the region of Yunnan [ https://en.wikipedia.org/wiki/Yunnan ] and is unrelated to Yunan.
23.
by a11r 4y ago
It depends. Sometimes you don't know that a solution exists for your problem.
24.
by a11r 4y ago
I am in the middle of listening to Einstein: His Life and Universe by Walter Isaacson and would highly recommend this biography to anyone interested in more.
25.
by a11r 4y ago
The State Controller's office publishes detailed reports: https://www.sco.ca.gov/ard_state_acfr.html
26.
by a11r 5y ago
Perhaps search for a library that has it. https://www.worldcat.org/title/last-empire-of-iran/oclc/1138...
27.
by a11r 5y ago
I think two mirrors at an angle would provide an opportunity to get much better precision in estimating the location on the screen being touched. Wondering if you have tried that.
28.
by a11r 5y ago
This is indeed the right way to leverage TCP flow control, and works well at the layer that knows exactly what is coming over the socket. The pitfall here is if semantically independent data is being multiplexed over the same TCP connection
29.
by a11r 5y ago
Even more incredible since the article points out that "Unlike albatross or other long-flying seabirds, godwits are active flyers, not gliders—their wings are moving the whole time." I wonder if they can use the active-flying equi
30.
by a11r 5y ago
The original paper by Burton H. Bloom is only 5 pages and very readable. https://dl.acm.org/doi/10.1145/362686.362692

More ›