Miyamoto Musashi @solminded

developer/builder. My twitter alter-ego for AI/LLM related stuff. Joined December 2021

Tweets

1K
Followers

172
Following

899
Likes

2K

Miyamoto Musashi @solminded

4 days ago

@FishAudio DRAMA

0 0 0 12 0

View Details

Miyamoto Musashi @solminded

a week ago

@JorCru With this logic the cover bands are all thiefs?

1 0 1 2K 0

View Details

Miyamoto Musashi @solminded

2 weeks ago

@notkevinzhang @typesafeai @CompleteSkeptic I have posted my use case under the thread yesterday. Fingers crossed.

0 0 0 34 0

View Details

@notkevinzhang I’d love to test it as a real-time control plane for AI agents on my agentic swarm framework. handling model routing, tool selection, reasoning effort, context filtering, and step-by-step verification to see how much latency and cost it can remove without sacrificing task quality

0 0 0 12 0

View Details

Miyamoto Musashi @solminded

2 weeks ago

Jev is really interesting because it pushes semantic intelligence toward the timescale of continuous, cheap decisions hundreds of milliseconds apart. This might revolutionize fields like trading, cybersecurity,moderation, ai agents, robotics and many more x.com/CompleteSkepti…

Diogo Almeida @CompleteSkeptic

2 weeks ago

4K 8K 76K 39.7M 59K

0 0 0 44 0

View Details

Miyamoto Musashi @solminded

2 weeks ago

@CompleteSkeptic This has great potential for autonomous agentic operations, cybersecurity, fraud/risk, algorithmic trading, live interaction/moderation and many other areas. I am now going over use cases at docs.typesafe.ai/concepts/use-c… Joined the waitlist

0 0 0 3 0

View Details

Diogo Almeida @CompleteSkeptic

2 weeks ago

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution

4K 8K 76K 39.7M 59K

View Details

Miyamoto Musashi @solminded

2 weeks ago

@NotSeanyTsunami Happy bday Sean!

1 0 1 18 0

View Details

Miyamoto Musashi @solminded

3 weeks ago

@YaoZhangX for $0.42 you can rent a rtx6000 pro for 25-30 minutes and can get pretty much the same performance with sol attention and easycache - and generate as many clips you like within the timeframe. this does not sound like a bargain.

0 0 1 26 0

View Details

Miyamoto Musashi @solminded

2 months ago

@alexcooldev Not game development, browser based simple game demos becoming fast.

0 0 0 52 0

View Details

Miyamoto Musashi @solminded

2 months ago

@kimmonismus What is the ROI on the $423 spent to make a browser based game demo? Asking for a friend.

0 0 3 673 0

View Details

Miyamoto Musashi @solminded

2 months ago

@doublenickk Any proof of that Marvel $5M buyout?

0 0 3 661 0

View Details

Miyamoto Musashi @solminded

2 months ago

@thsottiaux @blueemi99 let’s keep it removed please

0 0 1 125 0

View Details

Miyamoto Musashi @solminded

2 months ago

@satyanadella are you sure that skill is called “drill-me” not “grill-me”?

0 0 1 1K 0

View Details

Miyamoto Musashi @solminded

2 months ago

@thsottiaux Do not bring back 5h limits.

0 0 1 11 0

View Details

Miyamoto Musashi @solminded

2 months ago

@maxjendrall can you ssh into it?

2 0 3 3K 1

View Details

Miyamoto Musashi @solminded

2 months ago

@venturetwins @thinkymachines You could have used podcli or SupoClip github.com/FujiwaraChoki/… github.com/nmbrthirteen/p…

1 0 5 838 17

View Details

Miyamoto Musashi @solminded

2 months ago

@bakkermichiel It might be benchmark targeted training. Only hands on experience can tell whether it is valid or not.

0 0 0 143 0

View Details

Miyamoto Musashi @solminded

2 months ago

@theo I use Luna xhigh for grunt code implementation work. If you can assign sol/fable as a senior planner+reviewer closely watching the development pipeline it is a pretty cost effective solution.

1 0 0 298 0

View Details

Miyamoto Musashi @solminded

3 months ago

@davis7 I think this post has a valid argument against 5.6 x.com/sumukx/status/… In my experience your workflow needs to have a prevention mechanism to deal with the OCDness of 5.6 otherwise it goes overboard

Sumuk @sumukx

3 months ago

Some thoughts about GPT-5.6-Sol after ~30B tokens: Sol is the most OCD model I’ve used thus far. It very frequently gets one-shotted by random nits in the codebase and writes a bunch of tests to fix it. Even with fast mode, it’s incredibly slow to do this kind of iterative