Eric Choi

co-founder & cto @ (ly '25)

[Eric Choi @0xEricChoi]

@sariabuhamad @Probat_AI Huge

[Eric Choi @0xEricChoi]

@sambeskind Great runs!

[Eric Choi @0xEricChoi]

@sariabuhamad 🔥

[Eric Choi @0xEricChoi]

more startups have been born from this image than from talking to customers

[Eric Choi @0xEricChoi]

@sambeskind @dlevine815 @thechangj @krinetix1234 @ryanm_leung @PranitAgrawal4 @tej_thambi @kaushikm_ it was awesome hooping with everyone

[Eric Choi @0xEricChoi]

just felt like more people should know about deepwiki (from @cognition)

any github repo → instant docs you can actually chat with for free.

swap "https://t.co/jg7gMd1Rky" for "https://t.co/eVyYyfP8Mb" in any repo url and try it

[Eric Choi @0xEricChoi]

Check it out at https://t.co/Dv1ISRRTex

[Eric Choi @0xEricChoi]

Our landing page figures now obey gravity.
Details make UI feel real.

#Frontend #React #WebDesign

[Sari Abu-Hamad @sariabuhamad]

Spent the day at @ecomnorthsummit in LA. Main takeaway was to replace gut calls with data, but never lose the story.

Here's what I'm taking back 🧵

[Eric Choi @0xEricChoi]

@janaksunil Awesome work!

[Sari Abu-Hamad @sariabuhamad]

At #KBOS this week with @0xEricChoi , Sept 9-10 at the Hynes.

We build landing page personalization at Probat AI and prove the lift with holdouts. If you're wrestling with attribution, come find us. If you're not, come find us anyway.

Stop us in the hallway or send a note.

[Eric Choi @0xEricChoi]

@sariabuhamad Bang

[Eric Choi @0xEricChoi]

@avikrishna this is AGI

[Eric Choi @0xEricChoi]

@robzolkos share the interesting parts of the plan?

[Eric Choi @0xEricChoi]

@sariabuhamad Makes you understand why they’re the leading AI lab

[Eric Choi @0xEricChoi]

Anthropic just published the most detailed guide I've seen on building commerce agents, and the article is packed with useful information.

The recommended design rests on a standard agent loop (ik suprising).

I have summarized my major takeaways 🧵

[Eric Choi @0xEricChoi]

6/ This is the part that's applicable for all domains: the agent harness is what matters at the end of the day. The central model should always be a config change.

Read the article in greater detail here: https://t.co/YX2KaYTosD

[Eric Choi @0xEricChoi]

5/ Bullet points on production:
I. Write memory async in a separate process (13% higher fact recall than a save tool)
II. Models should stage; Person / Policy should apply
III. Grade snapshots, not conversations when running evals