Anhad Chawla

Founder - @Jobblerai , Co-Founder - SkinC , AI Enginneer - GDSC
I will motivate you , you motivate me

[Anhad Chawla @AnhadSChawla]

What if we translate each word in our prompt to a different language where that word is shorter in length?
Possible using a simple py script imagine how short our paragraph could get cause different languages are giving the shortest word.
So AI understands all languages so will it lower the input tokens it takes to formalize our prompt into machine code?

[Anhad Chawla @AnhadSChawla]

@pcshipp even if you do both your still homeless

[Anhad Chawla @AnhadSChawla]

Bruh wtf , wasn't claude code released last year!
Is the interview process gonna take 2 years?

[Anhad Chawla @AnhadSChawla]

you're not behind. you're just comparing your day 1 to someone's year 3 highlight reel.

[Anhad Chawla @AnhadSChawla]

5 signs you're on the wrong path:

save this

[Anhad Chawla @AnhadSChawla]

you don't need a ferrari to go to the grocery store.

if you're building anything with llms right now and you haven't audited your api costs, you're probably burning money for no reason.

most early builders make the same mistake:

they use the most expensive model for every task.

classification? gpt-4 tier.
summarization? gpt-4 tier.
simple routing? gpt-4 tier.

here's what i'd do this week:

1. list every place you call an api
2. tag each call: simple / medium / complex
3. route simple tasks to cheaper models
4. test locally first before burning credits in prod
5. compare cost, speed, and context limits side by side

before you lock in anything, look at real benchmarks:

https://t.co/B9tKl6af5w

and if you're testing agent logic, run it locally before you pay per token:

https://t.co/tz613JTpZQ

same architecture. way cheaper debugging.

this one change can literally save you hundreds a month before you even have revenue.

[Jobbler @Jobblerai]

why job searching in 2026 is broken (and what we built to fix it):

🧵

[Anhad Chawla @AnhadSChawla]

@anupamrjp Weird opinions....

[Anhad Chawla @AnhadSChawla]

@robj3d3 What about the fitbit air?

[Anhad Chawla @AnhadSChawla]

heard a founder in sf say they're "pre-revenue but post-vision."

bro you are unemployed with a figma file. just say that.

[Jobbler @Jobblerai]

applicants use ai to write resumes
recruiters use ai to reject those resumes
hiring is just bots fighting bots now

[Anhad Chawla @AnhadSChawla]

if your worker agent loops and you are still using console.log, you are guessing.

you need to see which tool call failed, what the payload was, and where context dropped.

wire up langsmith and watch the full timeline.

https://t.co/SZuYZAeOmn

[Jobbler @Jobblerai]

every company complains they cannot find good developers while simultaneously forcing applicants to create a new workday account and manually re-type their pdf resume into 45 different text boxes. the talent is there. they just closed the tab.

[Anhad Chawla @AnhadSChawla]

burning hundreds of dollars on api credits just to test a broken agent loop is insane.

if you are building orchestrators, test your logic locally first.

1. install ollama.
2. run ollama run llama3 in your terminal.
3. change the base url in your code to localhost:11434.

the routing logic remains exactly the same, but the debugging is free. swap the api key back to your premium model only when you push to production. get it here: https://t.co/iep8tHLSVN