Ketan Agrawal

technologist with an eye towards complex systems. building @usecarson (YC W26.) prev: snowflake, robust intelligence, stanford

[Ketan Agrawal @_ketan0]

@prathyvsh I’m not sure I understand this point!

I personally think there’s a lot of untapped alpha in creating interfaces (using AI) for understanding systems (physical systems, code systems, etc)…i feel this cuts in an anti-slop direction actually

[Ketan Agrawal @_ketan0]

https://t.co/Vd3NpnMluL

Lots of food for thought, & very ahead of its time. As a system designer you want to immerse yourself in the system, effortlessly scale levels of abstraction, form deep intuitions about its dynamics by playing with it

[Ketan Agrawal @_ketan0]

@prathyvsh Ah yeah, that talk is a classic

It’s crazy that ~half a century on, much of those ideas from Engelbart, Licklider, Ted Nelson etc. haven’t been fully realized

[Ketan Agrawal @_ketan0]

The awesome thing is that now (vs when @worrydream wrote this), creating these perspectives into your system is essentially free

[Ketan Agrawal @_ketan0]

RSI for me but not for thee

[Nabeel S. Qureshi @nabeelqu]

Interesting tidbit from the Mythos/Fable system card: Anthropic are invisibly nerfing any requests that target frontier LLM development.

[Ketan Agrawal @_ketan0]

'ask in side chat' is a nice affordance

[Ketan Agrawal @_ketan0]

@irl_danB @waitbutwhy Ahh I see, thanks… I definitely clocked the WBW life paths aesthetic but was curious if you’d had some slick way of making these

[Ketan Agrawal @_ketan0]

@TalbertHerndon 🔥 congrats!!

[Ketan Agrawal @_ketan0]

the human-AI experiences we design are limited by the metaphors we have to imagine with

“human in the loop” -> all anyone can imagine is an approval prompt

[a16z @a16z]

Mira Murati says frontier AI should be built like a tandem bike:

"Having humans in the loop doesn't quite describe it because it sounds like a checkpoint where we're signing off something, and then you're good to go."

"It's more like creating systems that are not just autonomously advancing and leaving civilization behind, but are more like a tandem bike."

"When you're going up a hill, maybe whoever is stronger is pedaling harder. But both hands are on the wheel. That's quite important because that's a different system. It's a system designed for collaboration."

"It will increase the level of agency that people have, and also it will help us steer the research direction towards creating outputs that are more value-aligned."

@miramurati at Bloomberg Tech live with @emilychangtv

[Ketan Agrawal @_ketan0]

@human_named_gui initially wispr’ed to it for 2+ hours to set the seed, brain dumped everything I could think of in my life

now providing it daily updates on my life, weekly reflections (messages in discord)

[Ketan Agrawal @_ketan0]

@badlogicgames I’m having a lot of fun as a 64gb m5 max owner

[Ketan Agrawal @_ketan0]

@erikdunteman Let’s go Erik!! So hype

[Ketan Agrawal @_ketan0]

@erikdunteman One other nice thing about harness outside the sandbox- you can 'lazy load' the sandbox, deferring to when the agent uses its first sandbox-requiring tool (read, write, bash, etc.) In the meantime, start streaming agent tokens to the user immediately

[George @odysseus0z]

https://t.co/GqhyL4KRJw

[Ketan Agrawal @_ketan0]

wow minimum viable ontology is such a cool idea

whether you’re writing lines of code or english, you need the vocabulary to be able to ask for what you want

[melody kim @melodyskim]

This tweet had me thinking, what’s the “minimum viable ontology” or list of terms to quickly get situated within a new domain, and improve your prompts?

Vibecoded https://t.co/IQS0grHbB5 to showcase this idea on a few example domains -- been wading into 3D recently and found this initial "domain map" helpful.

SO much to improve here but wanted to get this initial idea out! Would love thoughts / feedback :)

[Ketan Agrawal @_ketan0]

it takes a bit of time to get your sea legs with managing multiple agents at once. but eventually you start to figure it out

[Ketan Agrawal @_ketan0]

super excited to share what we've been working on! been a blast building with this team :)

[Y Combinator @ycombinator]

AI agents are trapped in chat bubbles. @usecarson breaks free by generating custom, task specific interfaces with enterprise-grade security.

Sign up today at https://t.co/JXVGTwfA0a.

Congrats on the launch, @sidharthkmenon, @_ketan0, @alex_iansiti, and @milanb_123!

https://t.co/lbJRHR4mYu

[Ketan Agrawal @_ketan0]

@mechanical_monk ‘always use context7 mcp to fetch docs before building’ kinda helps (but sometimes they still ignore the docs)