Tejas Kumar

@brooke_jamieson @Jonnymagic00 wonderful!! congrats you two!!

1 year married to @Jonnymagic00 today! 💒

@zeeg is this in the basement?

Progress

@imjonjia thanks!!!

非常精彩的讲解，basically series

@o_codes_life astaghfirullah

@TejasKumar_ What you say makes absolutely no sense

hear me out: strava but for books

@borna_perak no the kangaroo people

@TejasKumar_ You mean - Austrians, right?

just found out that australians consider schnitzel to be “australian classic food” ???????

@JoshuaNewton I made a version of this for one (or 66) specific book/s and it works quite well. dewfall.app It’s quite gamified and addictive.

@TejasKumar_ Dude, I’m really interested in this space. Would love to help in any way I can!

@th_mdo @manuelnavid no, its because goodreads' ux is ass

@TejasKumar_ @manuelnavid i think you've got your answer right here

@ddialar im building something, ill show you

@TejasKumar_ What would be the measure unit and how could we ensure the provided data are real and not fake incomes such the booktubers that say to read multiple books per month but they only talk about AI summaries?

@eugolive have you tried

@TejasKumar_ I don't think you can ride a book.

@manuelnavid no one i know uses this

@TejasKumar_ Isn’t that good reads??

i got asked today about the intersection between an agent harness and agent skills. here's how they work: - harness level: each skill's name, description, and "reference" is included in the system prompt and the agent is instructed to use when appropriate - agent level: skills are lazy loaded (tool call, `load_skill`) on demand when an agent recognizes it needs the skill, reads the skill's markdown content, and learns how to use associated tools with skill hope that clarifies and is helpful!

@viraj_kini @aiDotEngineer @varshagorrepati @manuelnavid I’m sure the team has a publishing plan.

@TejasKumar_ @aiDotEngineer @varshagorrepati @manuelnavid Where is the yt video ?

The slides + repo from my recent talk at @aidotengineer world’s fair in San Francisco are now available on GitHub. → github.com/TejasQ/basical… cc @varshagorrepati @manuelnavid

@Hem_chandiran you’re welcome my friend

This is a must for any AI agents. Very few are obsessed with evaluating them. 🥲 Without evals, an agent that works on your demo can quietly fail in production. The hard part isn't making an agent do something - it's proving it does the right thing, consistently.✅ Good evals help you 🎧: • Catch regressions after prompt/model changes. • Measure reliability instead of relying on vibes. • Compare models and agent strategies objectively. • Build confidence before shipping. Thanks for this goldmine @TejasKumar_ 👏

Frank Castle would never say “I guess so” like this. 😞

whenever this cursed app switches from “following” to “for you” automatically without my choosing, all i see is garbage

one of the keys to eating clean and being healthy is realizing that eating historically was never a routine, but a result: a result of a successful hunt. that’s exactly why eating and food itself is dopaminergic/rewarding (can we confirm, @hubermanlab?). understanding this helps rewire the reward circuitry to make exercise and clean eating also rewarding.