Nicholas Topousis

Interested in the intersection of ML and causal inference. also like steak. | he/him

[Nicholas Topousis @NickyTops]

Don’t feel great about this

[Nicholas Topousis @NickyTops]

@TheStalwart Not a full explanation, but writing code is how you learn how it works. The ongoing expectation for engineers is that they need to understand the code that they ship. If you're gonna need to take the time to review the code the AI wrote in detail, savings are limited.

[Nicholas Topousis @NickyTops]

@ashdgandhi Wdyt about LSE EME? Seems like the exception.

[Nicholas Topousis @NickyTops]

It is so sick that my cats both need to see a veterinary dermatologist

[Nicholas Topousis @NickyTops]

@Yuchenj_UW It’s an incentive issue not a skill issue. Most engineers can’t rattle off the regex to parse phone numbers. For a one off use case, it’s much easier to use ChatGPT. At the same time, engineers are incentivized to deliver quickly without being accountable for token usage.

[Nicholas Topousis @NickyTops]

@arpitrage Vs senior engineer and expert in Java trying to use an AI agent to make changes to an incredibly complex production, Java app that they wrote

[Nicholas Topousis @NickyTops]

@arpitrage At first blush, I think it’s a bit too early to write off AI impact on productivity. There is massive treatment effect heterogeneity at the individual and task level. e.g. Junior engineer learning how a GCP load balancer works by talking with ChatGPT …

[Nicholas Topousis @NickyTops]

@ONAN_OUS @bennpeifert UV

[Nicholas Topousis @NickyTops]

Currently using Cursor Agent, Perplexity, and ChatGPT to plan an insanely mid vaction for me and Ilana

[Nicholas Topousis @NickyTops]

@lionel_trolling It is if you want to solve the problem of “there are not enough grocery stores.” It is not if you want to solve the problem of “we want to make groceries cheaper.”

[Nicholas Topousis @NickyTops]

@21BDP21 @robirwinengr @optomachina If you’re making 216 base you’re probably getting to $350k after equity and bonus. All of those numbers in the original post included equity and (perhaps) bonus

[Nicholas Topousis @NickyTops]

@RexDouglass Hey! Let me know if you see anything at Block that stands out to you. Don’t think we have a ton of roles open but definitely interested in your quant expertise.

[Nicholas Topousis @NickyTops]

@ZaidJilani Link?

[Nicholas Topousis @NickyTops]

Y’all didn’t know about Shannon Divergence?

[UnHerd @unherd]

Claude Shannon laid the mathematical foundations of the digital world.

Yet he remains almost completely unknown. @Daniel_Kalder👇https://t.co/tVeKwSitas

[Nicholas Topousis @NickyTops]

@arnoblalam @RVrijj @Apoorva__Lal No!

[Nicholas Topousis @NickyTops]

@alz_zyd_ High end Chinese food is taking off in NYC

[Nicholas Topousis @NickyTops]

@alz_zyd_ It’s not just eating out. It’s living in a high CoL metro with:
1. No roommates
2. Some travel (e.g. one largish vacation and smaller trips interspersed throughout the year)
3. Eating out 2x per week
4. Grabbing a couple drinks with friends at a bar
5. Sufficient savings

[Nicholas Topousis @NickyTops]

@quantian1 Is this via cursor or just the chat interface?

[Nicholas Topousis @NickyTops]

@jeremyphoward This is true much more broadly. ML folks in industry need to remember that performance metrics (AUC, AUPR, etc) are used because they are generally correlated with business value which is a very specific function of your predictions, outcomes, and the business logic used

[Nicholas Topousis @NickyTops]

@RexDouglass Thousands of engineers and data professionals go to work every day and write not shitty code and commit to a repo after it passes peer review.

Every single line of code I write is reviewed by a peer before it’s merged to production.

[Nicholas Topousis @NickyTops]

@RexDouglass I am reading this and the replies from econ professors who refuse to learn git and take the time to write high quality code are making me crazy.