PhilippGille

6 ms·

1.
by PhilippGille 11d ago
> Open source does not apply to AI Isn't that a bit overgeneralized? There's more than weights for the Olmo models for example: https://allenai.org/olmo Similar for Nvidia's Nemotron models IIRC. Artificia
2.
by PhilippGille 19d ago
Token usage is generally lower in greenfield projects.
3.
by PhilippGille 21d ago
NovitaAI is already hosting it: https://openrouter.ai/z-ai/glm-5.3-flash#providers
4.
The AI boom: rational enthusiasm or the next dot-com bubble?(ecb.europa.eu)
3 points by PhilippGille 26d ago|1 comments
5.
by PhilippGille 1mo ago
Official blog post: https://blogs.nvidia.com/blog/geforce-now-thursday-linux-nat...
6.
by PhilippGille 1mo ago
According to OpenRouter, DeepSeek trains on input. At least when disabling all providers that train on your data, DeepSeek gets disabled. Other providers don't. You can also choose to route to ZDR only. https://openrouter.ai
7.
by PhilippGille 1mo ago
Existing recent discussion with 127 points, 68 comments: https://news.ycombinator.com/item?id=49246704
8.
by PhilippGille 1mo ago
98.33 according to https://mrshu.github.io/github-statuses/
9.
by PhilippGille 2mo ago
Yes that's my point. The old and the new version are different in capabilities, but now when someone talks about DeepSeek V4 Flash (in benchmarks, on inference providers), you don't know which exact version it's about. Some p
10.
by PhilippGille 2mo ago
That's what I mean. On DeepSeek it's now just `deepseek-v4-flash`, while OpenRouter calls it `deepseek/deepseek-v4-flash-0731`, so now when someone talks about DeepSeek V4 Flash, like in benchmarks, or other inference provide
11.
by PhilippGille 2mo ago
The previous V4 version wasn't called “Preview” by most inference providers. For example, the OpenRouter model slug was `deepseek/deepseek-v4-flash`. So now there will be confusion when someone talks about V4 Flash or when someone
12.
by PhilippGille 2mo ago
Depends on the reasoning effort, see https://deepswe.datacurve.ai (add Luna via model selection drop down, if it's not shown by default)
13.
Kagi releases official Kagi Assistant apps(kagi.com)
35 points by PhilippGille 2mo ago|1 comments
14.
New Framework Desktop Option with AMD Ryzen AI Max+ Pro 495 and 192GB Memory(frame.work)
79 points by PhilippGille 2mo ago|114 comments
15.
Cursor, Codex, Gemini CLI, Antigravity hit by sandbox escapes(bleepingcomputer.com)
2 points by PhilippGille 2mo ago|0 comments
16.
by PhilippGille 2mo ago
The project looks very interesting, thanks for sharing! You seem to have created a new GitHub account just for this project a week ago. Do you have any other GitHub accounts that enable us to see a track record of your work (maintenance, se
17.
by PhilippGille 2mo ago
Handy already supports streaming transcription models, and you can see the words in the small Handy pop-up while you are talking. So in general this definitely works. Handy is just missing the feature to insert these streamed words into the
18.
by PhilippGille 2mo ago
That was the case for early models (Llama etc), but they got much better since then. Not perfect, but good enough. This is from Ministral 3 14B, a 2025 model without reasoning, that you can run on your PC: > Write a Haiku involving Hacke
19.
by PhilippGille 3mo ago
Currently this is for payments with stablecoins. For Bitcoin / Lightning these kind of pay-per-request API paywalls have existed for many years already (e.g. my own from 8 years ago [1], but others as well). Flattr [2] existed for non-
20.
by PhilippGille 3mo ago
> Kimi and GLM models have coined a new term: Thinkslop. > [...] > So for now I'm happy with just two models: GPT and DeepSeek. 1. DeepSeek V3.2, V4 Flash, V4 Pro, at high or max thinking, ... when recommending a model it sho
21.
by PhilippGille 3mo ago
The interesting bits on how they achieved it: > On the model side, we applied FP4 quantization > introduced DFlash, an efficient speculative decoding method based on block-level masked parallel prediction > On the system side, Tile
22.
by PhilippGille 4mo ago
The blog post has more info: https://www.minimax.io/blog/minimax-m3
23.
by PhilippGille 4mo ago
Do you mean MiMo V2 Flash? V2.5 doesn't have a Flash version.
24.
by PhilippGille 4mo ago
It's in the article: > HTTP also allows the DuckDB-Wasm distribution to speak Quack natively! So DuckDB running in a browser can e.g., directly connect to a DuckDB instance running in an EC2 server using Quack.
25.
by PhilippGille 4mo ago
Both the original Markdown spec [1] as well as CommonMark [2] clearly specify support for inline HTML. With that you can kind of get the best of both words depending on your use case. For the most parts you just write the regular Markdown h
26.
La Suite Docs v5.0.0 released(github.com)
4 points by PhilippGille 4mo ago|0 comments
27.
by PhilippGille 4mo ago
On max it uses more than twice as many tokens as on high when running the ArtificialAnalysis benchmark suite, and then it's indeed the model with the highest token usage (among the current top tier models). See the "Intelligence v
28.
by PhilippGille 5mo ago
Benchmarks only paint part of the picture, but it's still a decent place to start looking into recent models: https://huggingface.co/spaces/mteb/leaderboard
29.
by PhilippGille 5mo ago
When you say "Gemini", which exact model do you mean? You know there are several and they vary a lot in how capable they are? Pro 3.1 Preview, 2.5 Pro (their latest non-preview pro model), Flash 3 Preview, ... Same with GPT-5: Lat
30.
by PhilippGille 5mo ago
> C# [...] only really works properly in Windows What do you mean with this? Maybe you are thinking of the old ".NET Framework" runtime, which only runs on Windows? Nowadays there is ".NET Core" which runs on macOS an

More ›