Akash Parashar

for the love of the game | creatively technical

[Akash Parashar @itsakaashhh]

I built Kahani.
It is an interactive world of fables built using world models by @reactorworld and narrations using @FishAudio

Through these stories, the child can take whatever actions he/she wants and learn the morals as consequences of those actions.

It's like stepping into those old kahaniya/stories your parents used to tell you. But here, you can choose your actions and see reap the fruits of those actions

Built for the #inception2 hackathon organized by @thelaunchd @raina_sai @MaxMill06 and team.
Loved being a part of this!

@aoagents was with me during the toughest prompt battles with the world models, by helping me develop the UI, harness, context layer, all of that as fast as I could

Check it out here:
https://t.co/KbftkNvy0r

[Akash Parashar @itsakaashhh]

@Rishet11 @aoagents Can't imagine handling this many workers and resolving all the conflicts by myself 😭

[Akash Parashar @itsakaashhh]

I have come to realise that

- MOST of your time goes into reading about the tools, the the research behind the idea.
- SOME of your time goes into actually creating the .MD files which your agent will use (or @aoagents agent orchestrator in my case)

This should only be for hackathons tho imo.
Would I prefer the pre AI hackathons era? Yes.

Because now there's very little learning incentive when the agent can handle the implementation flawlessly most of the time while you can watch shows or make your fav recipe or draw something, etc.
it's like everything has it's own pros and cons. shocker, right?

[Akash Parashar @itsakaashhh]

OCCAM

build in progress...
We gonna cook or we getting cooked ( w @marky_2813 )

@aoagents @maximor_ai @Tensormux @neatlogs
#syndicate

[Akash Parashar @itsakaashhh]

@marky_2813 @aoagents @maximor_ai @Tensormux @neatlogs the orchestrator works so I can enjoy my egg mayo sandwich in the meantime

[Akash Parashar @itsakaashhh]

OCCAM

build in progress...
We gonna cook or we getting cooked ( w @marky_2813 )

@aoagents @maximor_ai @Tensormux @neatlogs
#syndicate

[Kiki @laziedev]

after thinking about this for 4 days and finishing it right before submissions closed,

presenting ReturnSplit
- my Track 04 submission for the @Razorpay AI Buildathon 2026
what i built: https://t.co/bi5yAuoC2D

drop your honest reviews. don't sugarcoat them.

[Akash Parashar @itsakaashhh]

@laziedev @Razorpay really cool idea and editing 😼

[Akash Parashar @itsakaashhh]

It's a issue I was facing with @aoagents as well. My fedora just hanged and I had to forcefully shutdown and restart (32 gigs btw)

I found that the threads/workers which become merged and are left there instead of being terminated/archived take up a large amount of memory. Like 300-400mb per opencode worker for me.
Since then I just made sure to delete the idle/merged workers.

Maybe there can be a better fix for this, will have to go a bit deeper into this imo

[Akash Parashar @itsakaashhh]

@StefanoGPT Plus, India, YES

[Akash Parashar @itsakaashhh]

@kartikktwt Watch a sitcom like ted lasso. You can use it as a reward system, watching an episode after completing a task or something

[Akash Parashar @itsakaashhh]

@nandini__bagga Explain pls. Is it like a positive self talk thing?

[Akash Parashar @itsakaashhh]

@damnvikram SEO agent? How does it do SEO?

[Akash Parashar @itsakaashhh]

@_Deva_Path excited for this one

[Himanshu Jorwal @_Deva_Path]

been working on a game solo for over a month. it's called Disorder, a top-down 2d survival game inspired by 20 Minutes Till Dawn. never touched Godot before this, learning as i go. still rough and far from done. would mean a lot if you check it out once i drop a playable build

[Akash Parashar @itsakaashhh]

@aoagents_hack @Maaztwts @agent_wrapper

[Akash Parashar @itsakaashhh]

@Maaztwts @agent_wrapper Really excited for the hackathon hehe

[Akash Parashar @itsakaashhh]

Learning so much about automatic multi agent systems (A-MAS) and how they are actually useless against human coded MAS or even single agent systems when it comes to reasoning tasks or even parallelizable tasks.

Reading research papers is probably one of the most helpful things you can do when it comes to a field like harness optimization where things change fast.

The tracks in this @aoagents and @maximor_ai hackathon have become a good reason for me to go deep into this.

[Akash Parashar @itsakaashhh]

Happy with the setup I was able to achieve here hehe

Currently I'm
- Reading on multi agent systems for the syndicate hack (@aoagents)
- setting up a local dictation tool on fedora
- watching ted lasso in between

[Akash Parashar @itsakaashhh]

@PsudoKit Which crop is most profitable this season? Asking for a friend

[Akash Parashar @itsakaashhh]

@Rishet11 Come a long way. I used to easily crack the rotational motion numericals, now they might crack me

[Akash Parashar @itsakaashhh]

When did we go from rotations per minute (RPM) to requests per minute (RPM) ???

[Akash Parashar @itsakaashhh]

This is what happens when you don't research before letting an LLM slop write an article for you and publish it to medium 😭

context: wrote a whole article about an upcoming hackathon by @aoagents and mistook ao for @aoTheComputer