In my world of Dreams and Hopes. There was a day of pain. There was a day of Struggle. There was a day of low . There was a day of nothing. But . . . . . . There was never a day without Dreaming.. Counting the hours and days .. #OneDay #Random #Life

The Ask feature in Photos is Soo helpful.

What's happening with $WSPP. for last few days its on pump..

Invested almost 300 memecoins...For last 18 months . Current standing: 75% loss

Models which can think better as we grow... I think first time i have seen this ... #Antigravity #VibeCoding.

If you are using @antigravity , try to use /goal and /grill-me, they are very usefull.

People or orgs who used mythos as part of Project Glasswing, did they sign any NDA?

People on the other side. There should be Trillionaire Tax.....

That's huge....

Will you buy the SpaceX IPO?

When Elon Reaches $1T. it would be ₹95,302,500,000,000 inr. Just count.....

A week ago I had a dumb question. Why am I paying to send my entire codebase to an LLM? Every new model announcement seems to be: "Now supports even more context!" But context isn't free. More tokens = more cost, more latency, more noise. So I started a small experiment. First I stripped comments. Then dead code. Then I asked: "What if I remove the implementation entirely and only keep the architecture?" That became LeanContext. In about a week I built: • A VS Code extension • An MCP server • A repository compression engine • A benchmarking framework The latest experiment is called Skeleton Mode. Instead of sending full source files, it keeps: imports/exports classes interfaces type definitions function signatures and removes method bodies. Results on real repositories: Raw Context: 667,992 tokens Minified: 646,770 tokens (-3.2%) Skeleton: 361,759 tokens (-45.8%) Then I ran a reasoning benchmark. Full Context: Correctness: 4.19/5 Reasoning: 4.45/5 Skeleton: Correctness: 3.90/5 Reasoning: 4.33/5 So far: • ~46% fewer tokens • ~46% lower cost • ~93% correctness retained • ~97% reasoning quality retained It's still early and the sample size is small. But the result surprised me. The useful information in a repository might not be the implementation. It might be the architecture. Next step: validate across more repositories and languages. Either the hypothesis survives, or it dies quickly. Both outcomes are useful.

Doing the work is different to Learning at work.

Things that should be more talked, more shared and more inspired.... Congratulations 🎉🎉🎉

From Great model at task building to Cheaper model to task building.... World changed alot.....

Did one more test on large set of files the result is looking promising at the moment, what started as exploration and show some skill set to get an interesting job. #OpenSource #LeanContext #MCP

So, if you got a chance to Test around #Fable. How's it? Do you like it?