Viewing profile — try-working
try-working
HN member- Joined
- Sun, Mar 15, 2026, 3:49 AM UTC
- HN karma
- 266
- Public activity
- 205 items
- HN profile
- View on Hacker News ↗
About try-working
https://x.com/trydotworks
model router: https://github.com/try-works/role-model
Recent public activity
-
comment
Comment #49220597
No, you still don't understand what I'm saying. Yes, ASICs make inference faster, but also makes the hardware obsolete faster, if it's embedded in a phone. That sounds like a negat…
-
comment
Comment #49215824
What I'm saying is that Apple will use these type of models etched into chips, and they will do it because it drives obsolescence, so they can shorten the upgrade cycle. They will …
-
comment
Comment #49206581
What's even more noticable is that Anthropic still hasn't responded to the Kimi K3 release or the DeepSeek release.
-
comment
Comment #49205774
you don't understand what I wrote.
-
comment
Comment #49203664
obsolescence is the whole point. apple gets to sell a new phone very 6-12 months because of it. i have written about this: "For device makers Packaging models with laptops and smar…
-
comment
Comment #49189487
Flash is the most used model in the world since last week
-
comment
Comment #49189476
yes, and this is why we need model routing
-
comment
Comment #49181062
it's 200 rmb and it looks super fun. perfect product.
-
comment
Comment #49167898
I recommend the Shadowrun series on Steam. Not large games but really fun.
-
comment
Comment #49147776
agree. what people dont understand is that the blacklisting is what created motivation for investing capital, manpower and years of time into these companies, or initiatives inside…
-
comment
Comment #49147744
this is not a good benchmark for models, but it's great if you're optimizing for attention on twitter because video content and 3d animations perform best on social media. a real b…
-
comment
Comment #49142874
No, you are wrong. Without the blacklisting and export bans the DUV machine in the article would not exist today.
-
comment
Comment #49140746
All of this can be traced back to the 2018 blacklisting of ZTE, and the continued willful ignorance of China's capabilities, and of the history of industrialization and technologic…
-
comment
Comment #49139132
that's why i recommend running everything through a router.
-
comment
Comment #49128546
the repo is here and you can also find my twitter in my profile: https://github.com/try-works/role-model you can also read this: https://try.works/first-principles-of-model-routing…
-
comment
Comment #49128490
yep. I've been saying for month that exactly this will happen with DeepSeek and Kimi K3, and margins will erode, and Anthropic will fail to IPO.
-
comment
Comment #49128486
yeah it's pretty cool. you can also use it as a regular music player. if you toggle autoplay, it will automatically keep playing the next track in your track list. supports uploadi…
-
comment
Comment #49128372
I recently wrote about first principles of model routing that I've learnt building a model router. The model pool should be kept small, and models in the pool should be clearly dif…
-
comment
Comment #49128288
I built something similar, but for the browser: https://dj.t-tunes.com/
-
comment
Comment #49122033
Now let's see Dario's price cut.
-
comment
Comment #49120481
pi-role-model
-
comment
Comment #49120320
[flagged]
-
comment
Comment #49120251
OpenAI is doing this so that you can't preserve cache when routing between for example GPT 5.6 and DeepSeek V4, or other providers. I'm building a router that easily reaches >99% c…
-
comment
Comment #49120170
Let's see how the market reacts.
-
comment
Comment #49120162
DeepSeek themselves called it `deepseek/deepseek-v4-flash`. Pro is still like that.