Eric Chen

building more things | prev @pennmandt

[Eric Chen @ChenOutOfChen]

The analytics have to be wrong. There’s no way sitting in a low block and eating crosses to the face is the most positive EV tactic to hold a 1-0 lead.

[Eric Chen @ChenOutOfChen]

@erikbzhang @PoeticHQ Congrats man!

[erikz @erikbzhang]

New chapter – I’ve left Palantir to join the team @poetichq!

Having spent the last year and a half immersed in enterprise AI, it became a no-brainer that Poetic is the team to bet on.

The thesis: the industry is inundated with flashy demos and shallow wins. Poetic is different: we’ve proven out 8 digits of savings in months, and this is just the beginning.

Most importantly, nobody does it like us. Through Sand (our own coding language/DSL), Poetic uses LLMs to compile half-code, half-English into mostly deterministic workflows, then runs them with as little inference as possible in production.

That makes the product faster, cheaper, and more reliable than using an agent at every step. And once SOPs, videos, and tribal knowledge are systemized as code running at the nines of accuracy, you can backtest, diff, and govern your runbook like software.

On a personal level, I’ve loved every single person I’ve met here. Every story I hear about why someone is here deepens my conviction. The level of intensity and talent at Poetic is deeply energizing.

We are now bottlenecked by talent. We’re hiring. Shoot me a DM:)

[erikz @erikbzhang]

I made Claude for Chrome up to 100% faster in a day with a simple concept.
Browser agents like Claude for Chrome are making way too many inference calls.
Most people use the same ~20 websites. The interaction patterns on those sites are learnable and reusable. So why hit the LLM every step?
I built a PoC where the LLM gets the task, picks the site, and pulls from a library of learned patterns for that site, like a cache. It plans what to do, then executes. No back-and-forth probing, no redundant inference calls to locate elements it found last week.
Results: up to 100% faster, ~25 fewer inference calls per task, half the cost, and sometimes more successful in actually getting the task done.

[Sourish Jasti @SourishJasti]

1/ The future of general-purpose robotics will be decided by one major question: which flavor of data scales reasoning? Every major lab represents a different bet.

Over the past 3 months, @adam_patni, @vriishin, and I read the core research papers, spoke with staff at the major labs, and mapped the talent pool. This has completely changed how we think about general-purpose robotics.

Our paper builds intuition, step-by step, across the 2025 frontier: from architectures → evals → data → industry dynamics. Each layer reveals a different bottleneck, but they all converge on one truth—data decides everything.

Our takeaways + process below👇

If you want access to our graph (sound on), comment or DM me

[Eric Chen @ChenOutOfChen]

@erikbzhang Could it be that your whole X feed is tech and finance influencers and your ig feed is gen z influencers…

[Eric Chen @ChenOutOfChen]

Does that mean if at least one of our degrees is still useful? 😭

[Ria Sharma @ririsharms]

We gave the top 4 LLMs 1,800 real accounting and valuation problems from MIT and Wharton.

Best score? 71%.

Not a single model would’ve passed a Goldman interview. Here’s why it matters 🧵(1/7)

[erikz @erikbzhang]

1/ I wrote this thesis a bit over a month ago and sent it to several TradFi money managers managing a total of hundreds of millions - it net 30% in one month.
The market is continuing to show that ETH is fundamentally mispriced in the crypto ecosystem (this is from someone who used to basically only own ETH).
Here's why the BTC/ETH spread trade is about to get much bigger (as much of CT knows) 🧵. Full 13 page pitch linked below.

[Eric Chen @ChenOutOfChen]

@senapathy__ that’s crazy, where’d u hear that

[Eric Chen @ChenOutOfChen]

@senapathy__ tony stark arc incoming

[Eric Chen @ChenOutOfChen]

Read the full article on our substack: https://t.co/AtinFj5mTC

[Eric Chen @ChenOutOfChen]

1/🌐 As #Web3 goes multi-chain, #interoperability becomes vital.

Will @Polkadot's soon-to-be-released XCMP or @Cosmos's IBC emerge as the dominant protocol?

In collaboration with @d1ventures, we try to answer this important question through a UX lens for end-users. #XCMP #IBC

[Eric Chen @ChenOutOfChen]

If you liked this thread, follow @franklin_dao for more crypto research and @d1ventures for what's hot in Web3. Thank you to @erikbzhang for the support!