Parijat Bhattacharjee

24,fullstack dev, shipping features building more

[Parijat Bhattacharjee @poborojo]

@adityab29_ What about opencode ?

[Parijat Bhattacharjee @poborojo]

@shrav_10 Add please

[Parijat Bhattacharjee @poborojo]

@OpenAI But where is 5.6 ?

[Rohan Paul @rohanpaul_ai]

🎯 Andrej Karpathy on how to learn.

[Parijat Bhattacharjee @poborojo]

@jgravelle1963 Okay got it thanks

[Parijat Bhattacharjee @poborojo]

1/ Been exploring MCP servers lately and stumbled across jCodeMunch. Honestly did not expect it to change how I think about giving context to LLMs. Here is what happened.

[Parijat Bhattacharjee @poborojo]

@jgravelle1963 Wanted to know 1 thing how does the jdocmunch works ? Does it looks up the latest documentations from the web like context7

[Parijat Bhattacharjee @poborojo]

This is the better way to understand long and boring markdowns, I used it and damn its really good, the markdown which would have taken me 20-30 mins to understand, just took me 5 mins with this trick.

[Thariq @trq212]

https://t.co/MXt5XS4xBX

[Shivang Yadav @16_shivang]

15k in Delhi vs 15k in Bangalore

[Parijat Bhattacharjee @poborojo]

Exactly I still feel the value 5.4 provides is far more better than claude models

[Theo - t3.gg @theo]

It is really funny to me that Sonnet is still more expensive than GPT-5.4

[Parijat Bhattacharjee @poborojo]

@nalinrajput23 It's better to use 5.4 tbh

[Karan @karankendre]

Kimi just released K2.6

>open sourced
>swe-bench multilingual 76.7
>handles large-scale coding tasks
>runs thousands of tool actions
>builds complex frontend systems
>supports devops + optimization
>one prompt can generate 100+ files

[Wise @trikcode]

So GitHub Copilot just paused new Pro signups
because they ran out of compute.

the biggest AI coding tool on the planet literally cannot handle the demand.

the AI infrastructure crisis is here

[Parijat Bhattacharjee @poborojo]

This is really big an OS model leading the AI race

[BridgeMind @bridgemindai]

Kimi K2.6 just dropped.

And it crushed Claude Opus 4.6 on SWE-Bench Pro.

Kimi K2.6: 58.6
GPT-5.4 xhigh: 57.7
Gemini 3.1 Pro: 54.2
Claude Opus 4.6: 53.4

An open source Chinese model is now #1 on agentic coding.

Frontier labs have a problem.

[Wes Winder @weswinder]

gpt-5.4 is still so much better than opus 4.7 and it’s not even close

[Parijat Bhattacharjee @poborojo]

@DhruvL515 It's not what it always looks like it's an awareness campaign

[Parijat Bhattacharjee @poborojo]

@Itstheanurag

[Stanlee | Web developer @stanlee0nX]

“npm install” :

[Parijat Bhattacharjee @poborojo]

@mehulmpt Got to see that's claude is 2x on weekends is it true ?

[Parijat Bhattacharjee @poborojo]

v1 kanha GPT: one prompt.
people liked it. but krishna was just improvising.

rebuilt it with RAG + vector.
every answer grounded in actual gita verses.
you can verify which shloka informed it.

v1 was an actor playing krishna.
v2 reads what krishna said.

https://t.co/iaUstNEw19

[Parijat Bhattacharjee @poborojo]

@TomCat415 @jgravelle1963 Trust me you won't regret it @jgravelle1963 did a really great job