JPawchannel 🚀
partially tped at 63.5 and added here also added gold btc thesis same. shiny rock thesis - idk chart looks gud ^_^ trd
btc so far looks gud. added more at 63.3
trd
trd
JPawchannel 🚀
fable vs gpt 5.6 sol benchs say fable better so am rly surprised to say that i think sol is better. usually slightly better, but sometimes significantly better am using fable as orchestrator and sol as worker. wen i ask fable a question, i tell to get answer…
feels like fable will not leave sub cuz without it anthropic's subscription will not worth having. even now i think it's not gud cuz usage gets drained way faster than on sol, plus sol is actually better. it's wiser to buy second openai sub than anthropic's
JPawchannel 🚀
we're so early in agentic ai era and pre agi era ai will gib biggest bubble eventually with a lot of smol bubbles on the way which will look like big bubble. for example recent capex narrative was telling ai bubble will soon burst and it did gib a bit of…
codex active users growing rapidly. i'd guess there's abt 15M active vibe coders globally. we're still v early
for any terra or gpt 5.5 effort level, there is a luna or sol effort level that is more intelligent at no extra cost, or as intelligent at lower cost
also noticed in my practice that sol often goes v deep into implementation and takes way more time without significant trade off in quality. hence now more than even am using lesser capable models for task delegation (for workers under orchestration) and sol and fable mostly for orchestration
also noticed in my practice that sol often goes v deep into implementation and takes way more time without significant trade off in quality. hence now more than even am using lesser capable models for task delegation (for workers under orchestration) and sol and fable mostly for orchestration
wow moonshot released kimi k3
benching coding at 76.2. above gpt 5.5 and opus 4.8. behind sol terra and fable
and they planning to open source it 😮 it wud be the best open weights model
2.8T params with 1M context window
we're almost at next stage of ai race wer open source will dominate in general purpose models. wud guess by q4 we'll see openai and anthropic pivoting to specialized models
won't be surprised if openai and anthropic start serving open models next year in parallel to specialized
benching coding at 76.2. above gpt 5.5 and opus 4.8. behind sol terra and fable
and they planning to open source it 😮 it wud be the best open weights model
2.8T params with 1M context window
we're almost at next stage of ai race wer open source will dominate in general purpose models. wud guess by q4 we'll see openai and anthropic pivoting to specialized models
won't be surprised if openai and anthropic start serving open models next year in parallel to specialized
JPawchannel 🚀
codex active users growing rapidly. i'd guess there's abt 15M active vibe coders globally. we're still v early
strong trend. now 2.2% of us households gave ai subscription
JPawchannel 🚀
btc so far looks gud. added more at 63.3 trd
took some profits. feels like we cud get one more dip to 63
also averaged down la and gram
trd positioning
also averaged down la and gram
trd positioning
Baton - Agentic Orchestration Framework
The Secret Behind Fable 5 Ultracode and GPT 5.6 Sol Ultra and How Baton Takes It Further
New Substack Post:
https://jpawchan.substack.com/p/baton-agentic-orchestration-framework
The Secret Behind Fable 5 Ultracode and GPT 5.6 Sol Ultra and How Baton Takes It Further
New Substack Post:
https://jpawchan.substack.com/p/baton-agentic-orchestration-framework
JPawchannel 🚀
took some profits. feels like we cud get one more dip to 63 also averaged down la and gram trd positioning
la looks like cup and handle in ltf, might push to 0.15 if it holds 0.6
ta trd positioning
ta trd positioning