Kentaromiura
@kentaromiura.bsky.social
Reposted by Kentaromiura
i'm looking for a new job
Hire Me in Japan — overreacted
I'm looking for a new job.
overreacted.io
November 10, 2025 at 3:31 PM
i'm looking for a new job
Some Tuesday at work we have a “vibe engineering” session where we discuss stuff, this Tuesday I spoke on how to make AI agents that use local llms to reproduce @roocode.bsky.social “boomerang” mode on the cli with multiple local llms using github.com/kentaromiura... and @lmstudio-ai.bsky.social
GitHub - kentaromiura/ODM: On demand model proxy server
On demand model proxy server. Contribute to kentaromiura/ODM development by creating an account on GitHub.
github.com
October 8, 2025 at 11:22 PM
Some Tuesday at work we have a “vibe engineering” session where we discuss stuff, this Tuesday I spoke on how to make AI agents that use local llms to reproduce @roocode.bsky.social “boomerang” mode on the cli with multiple local llms using github.com/kentaromiura... and @lmstudio-ai.bsky.social
Since this I’ve released two local tools already.
The first for Jp/En translation and back and the latest one for allowing to load and unload different models on demand, allowing better agentic flow on local development machine.
ODM: github.com/kentaromiura...
ShallowT: github.com/kentaromiura...
The first for Jp/En translation and back and the latest one for allowing to load and unload different models on demand, allowing better agentic flow on local development machine.
ODM: github.com/kentaromiura...
ShallowT: github.com/kentaromiura...
September 29, 2025 at 12:23 AM
Since this I’ve released two local tools already.
The first for Jp/En translation and back and the latest one for allowing to load and unload different models on demand, allowing better agentic flow on local development machine.
ODM: github.com/kentaromiura...
ShallowT: github.com/kentaromiura...
The first for Jp/En translation and back and the latest one for allowing to load and unload different models on demand, allowing better agentic flow on local development machine.
ODM: github.com/kentaromiura...
ShallowT: github.com/kentaromiura...
I’ve fully entered the local llm rabbit hole. Please send help (or an AMD AI MAX if you hate me).
August 17, 2025 at 11:28 PM
I’ve fully entered the local llm rabbit hole. Please send help (or an AMD AI MAX if you hate me).
This is the best advertisement for grok tbqh.
August 13, 2025 at 12:00 AM
This is the best advertisement for grok tbqh.
A few years ago (2011) I made a script that downloaded lots of news from different rss feeds, put together in a page, run through calibre to convert to mobi and send to my kindle every morning; I think this is now a prompt away from that.
So yesterday Jan V1 was released based on this model. So I tried Jan in the Jan app, after configuring the duck duck go and fetcher MCP I managed to get a little assistant to summarize news etc, although it got Wikipedia winning instead of losing.
Despite this, for the first time I’ve got 2 local llms finishing my test, both the stubborn but small qwen Qwen/Qwen3-4B-Thinking-2507 and gpt-oss-20b, the latter with huge context could even complete the task on first try via aider.
Test is asking to write rule 110 in Odin.
Test is asking to write rule 110 in Odin.
August 12, 2025 at 11:40 PM
A few years ago (2011) I made a script that downloaded lots of news from different rss feeds, put together in a page, run through calibre to convert to mobi and send to my kindle every morning; I think this is now a prompt away from that.
So yesterday Jan V1 was released based on this model. So I tried Jan in the Jan app, after configuring the duck duck go and fetcher MCP I managed to get a little assistant to summarize news etc, although it got Wikipedia winning instead of losing.
Despite this, for the first time I’ve got 2 local llms finishing my test, both the stubborn but small qwen Qwen/Qwen3-4B-Thinking-2507 and gpt-oss-20b, the latter with huge context could even complete the task on first try via aider.
Test is asking to write rule 110 in Odin.
Test is asking to write rule 110 in Odin.
I think AI is remarkably similar to human engineer behavior. Maybe a bit too much.
August 12, 2025 at 11:28 PM
So yesterday Jan V1 was released based on this model. So I tried Jan in the Jan app, after configuring the duck duck go and fetcher MCP I managed to get a little assistant to summarize news etc, although it got Wikipedia winning instead of losing.
Expedition 33 has one of the worst platforming part of any game. If I was to demo the game and give it a score solely on that it would get a 0.
August 10, 2025 at 6:12 AM
Expedition 33 has one of the worst platforming part of any game. If I was to demo the game and give it a score solely on that it would get a 0.
Despite this, for the first time I’ve got 2 local llms finishing my test, both the stubborn but small qwen Qwen/Qwen3-4B-Thinking-2507 and gpt-oss-20b, the latter with huge context could even complete the task on first try via aider.
Test is asking to write rule 110 in Odin.
Test is asking to write rule 110 in Odin.
I think AI is remarkably similar to human engineer behavior. Maybe a bit too much.
>Hey AI, I think instead of rule 110, you implemented rule 118, you’ve got the MSB and LSB backwards.
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
August 8, 2025 at 10:03 AM
Despite this, for the first time I’ve got 2 local llms finishing my test, both the stubborn but small qwen Qwen/Qwen3-4B-Thinking-2507 and gpt-oss-20b, the latter with huge context could even complete the task on first try via aider.
Test is asking to write rule 110 in Odin.
Test is asking to write rule 110 in Odin.
I think AI is remarkably similar to human engineer behavior. Maybe a bit too much.
>Hey AI, I think instead of rule 110, you implemented rule 118, you’ve got the MSB and LSB backwards.
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
August 8, 2025 at 9:51 AM
I think AI is remarkably similar to human engineer behavior. Maybe a bit too much.
>Hey AI, I think instead of rule 110, you implemented rule 118, you’ve got the MSB and LSB backwards.
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
August 8, 2025 at 9:45 AM
>Hey AI, I think instead of rule 110, you implemented rule 118, you’ve got the MSB and LSB backwards.
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
Ai: I see that’s a common mistake when implementing cellular automations, I’m actually correct.
>No you’re not;you are reading the truth table backwards, MSB is bit 7 you think is 0.
Ai: Acthually..
TIL if you zoom enough you can read the “Here’s to the crazy ones.” In the document emoji on a Mac.
August 5, 2025 at 3:53 AM
TIL if you zoom enough you can read the “Here’s to the crazy ones.” In the document emoji on a Mac.
So you all know Serial Experiment Lain, but do you know
August 4, 2025 at 12:39 PM
So you all know Serial Experiment Lain, but do you know
As usual, #TokyoGameDungeon was pretty good.
August 3, 2025 at 7:10 AM
As usual, #TokyoGameDungeon was pretty good.
Latest @actionretro.bsky.social video reminded me I actually have a GSA certification, and GSA were discontinued after 2018. Fun fact after buying one in my old company I spent 3 days building a proxy +replacement using lucene and a custom xslt transformation to replicate what we were using it for.
August 3, 2025 at 1:46 AM
Latest @actionretro.bsky.social video reminded me I actually have a GSA certification, and GSA were discontinued after 2018. Fun fact after buying one in my old company I spent 3 days building a proxy +replacement using lucene and a custom xslt transformation to replicate what we were using it for.
I cannot start my morning without a cup of ‘cino.
July 31, 2025 at 12:09 AM
I cannot start my morning without a cup of ‘cino.
After all it was one of the first to employ AI when making alien faster the less they were on the screen.
July 29, 2025 at 10:05 AM
After all it was one of the first to employ AI when making alien faster the less they were on the screen.
Consumer grade hardware. Local llm. This is super helpful in good hands. This is also devastating for the industry, it’s ATM an intern equivalent, the difference is that it can iterate way faster when it gets it wrong.
Using it feels both exciting and a punch in the gut.
Using it feels both exciting and a punch in the gut.
July 21, 2025 at 4:07 AM
Consumer grade hardware. Local llm. This is super helpful in good hands. This is also devastating for the industry, it’s ATM an intern equivalent, the difference is that it can iterate way faster when it gets it wrong.
Using it feels both exciting and a punch in the gut.
Using it feels both exciting and a punch in the gut.
This is my reply when people says: AI will make software developers* useless.
* also applicable to many other professions.
* also applicable to many other professions.
If you have to replace a wooden door you’ve 3 options, you can either diy with the help of some yt videos, and regret all the mistakes you’ll eventually do, ask your friend that likes to do bricolage as an hobby if he can help and he might be able to help or not, or call a professional.
July 14, 2025 at 12:37 PM
This is my reply when people says: AI will make software developers* useless.
* also applicable to many other professions.
* also applicable to many other professions.
If you have to replace a wooden door you’ve 3 options, you can either diy with the help of some yt videos, and regret all the mistakes you’ll eventually do, ask your friend that likes to do bricolage as an hobby if he can help and he might be able to help or not, or call a professional.
July 14, 2025 at 12:34 PM
If you have to replace a wooden door you’ve 3 options, you can either diy with the help of some yt videos, and regret all the mistakes you’ll eventually do, ask your friend that likes to do bricolage as an hobby if he can help and he might be able to help or not, or call a professional.
Watching a @bradleone.bsky.social video on YouTube when an ad interrupts in the middle of a speech with no pause in between:
Brad: <>
Grammarly ad:<>
Nicely done algorithm, nicely done.
Brad: <>
Grammarly ad:<
Nicely done algorithm, nicely done.
July 5, 2025 at 4:25 AM
Watching a @bradleone.bsky.social video on YouTube when an ad interrupts in the middle of a speech with no pause in between:
Brad: <>
Grammarly ad:<>
Nicely done algorithm, nicely done.
Brad: <>
Grammarly ad:<
Nicely done algorithm, nicely done.
Why THE FUCK YouTube now starts videos with automatic robot translation and gives no option to switch to real track, what the fuck is wrong with you YouTube?
July 2, 2025 at 12:18 AM
Why THE FUCK YouTube now starts videos with automatic robot translation and gives no option to switch to real track, what the fuck is wrong with you YouTube?
I read about github.com/Austin1serb/... on hacker news and I do something very similar here github.com/kentaromiura..., only I use css variables instead of data attributes on the body, it keep UI extremely fast, at 100% score on lighthouse. Maybe I should write a post about it.
GitHub - Austin1serb/React-Zero-UI
Contribute to Austin1serb/React-Zero-UI development by creating an account on GitHub.
github.com
June 21, 2025 at 6:11 AM
I read about github.com/Austin1serb/... on hacker news and I do something very similar here github.com/kentaromiura..., only I use css variables instead of data attributes on the body, it keep UI extremely fast, at 100% score on lighthouse. Maybe I should write a post about it.