@[email protected]

#insurtech consultant researcher & community advocate—& inveterate data nerd. Chief Data Officer @ #RStats lover, #Python enthusiast, writer.

[@[email protected] @BenjaminWolfe]

I'm finally ready to share the new chapter of my career! 🎉

I joined a small business, https://t.co/VebxEshfAe, as an #insurtech consultant, researcher & community advocate.

I'm also fortunate to be Finsure's Chief Data Officer, so it's not like I'm not doing data work!

🧵1/10

[@[email protected] @BenjaminWolfe]

@tanya_shapiro Can I just say

I'm almost never on this app nowadays; it's hard to get over what the new overlord has done with it.

But your content is consistently interesting and quality enough to have me actually tap those notifications!

[@[email protected] @BenjaminWolfe]

Oh no how do I spell Business Owners Policy properly? I may never know. @TheHartford I somehow thought you might be definitive. 😅

@insurtechATX can my hometown folks come to my rescue here?

#apostrophe #insurance #insurtech #style

[@[email protected] @BenjaminWolfe]

If you see this - quote tweet with the last picture of your pet - no cheating

(Why not.)

[@[email protected] @BenjaminWolfe]

@tanya_shapiro Oh my. My Peloton plot would have a lot more white on it. 😅😅😅

[@[email protected] @BenjaminWolfe]

This is a fascinating, unexpected thread. There are weird things I would not have expected to make a difference in #ChatGPT prompts.

[Peter Yang @petergyang]

3/ Help the LLM succeed

An LLM has no notion of success. You have to make it want to succeed with prompts like:

• "You're an expert in X"
• "Think through this step by step"
• "Make sure you have the right answer"

Here's how the above helps with accuracy:

[@[email protected] @BenjaminWolfe]

@CMastication I often say rubber-ducking doesn't quite work with an actual rubber duck.

For me, the kind of minimum is that I believe I'm talking to a mind capable of understanding my problem.

And I mean, if you listen to like EVERY song about Jesus…

[@[email protected] @BenjaminWolfe]

@CMastication Daaaaaamn that's deep. Damn.

[@[email protected] @BenjaminWolfe]

@TorresOnTech @github Yes!! Kind of an evergreen reminder.

[@[email protected] @BenjaminWolfe]

Debugging tip: Go for a walk 🐕‍🦺🚶

I pushed a partial commit, copy-pasted the stack trace to my email, and walked off with my dog and my phone. (thx @github app…)

Found like 2 bugs and 3 more edge cases. Sometimes what needs a reboot is your brain. 🧠

[@[email protected] @BenjaminWolfe]

@charliejhadley Also you totally didn't come across passive aggressive! I can't even think how to interpret it that way. 😅

I was just imagining how ChatGPT might answer it.

[@[email protected] @BenjaminWolfe]

Been writing a bunch of code without logging. Oops.

No problem for ChatGPT!

"Can you write a bunch of print statements throughout this code so I can see what's happening when it runs?"

✨magic✨

[@[email protected] @BenjaminWolfe]

@charliejhadley Yeah I literally say Thanks and stuff, like she's Rosie from The Jetsons.

[@[email protected] @BenjaminWolfe]

unsolicited VSCode tip…

I'm finding "split right" is a *really* useful feature.

(You right-click on a tab and click "split right" to see the same file you're looking at in a second pane.)

You can be working in line 1000 of a long file and glance at line 7 without scrolling.🤔

[@[email protected] @BenjaminWolfe]

Holy smokes! This is awesome. @pydantic is what @FastAPI uses under the hood. And FastAPI is a really fun #python framework to code APIs with.

[Sebastián Ramírez @tiangolo]

How is @pydantic v2 rewritten in Rust and 22x faster (that's 2200% faster) by @samuelcolvin 😎🚀🐍

[@[email protected] @BenjaminWolfe]

(?<=xyz) isn't a capture group. It means the uppercase letter you're capturing should be preceded by xyz.

$1 in the "replace" means use that capture group. If you had two you might use $2, etc.

\l means lowercase it! (That's an L.)

_ means add an underscore. :)

3/3

[@[email protected] @BenjaminWolfe]

Did you know VSCode can not only do regex replacements, but change case in the process?

I had a bunch of two-word camelCase variables and had to change them to snake_case. They always came after a certain string.

Find: (?<=a certain string [a-z]+)([A-Z])
Replace: _\l$1

/1

[@[email protected] @BenjaminWolfe]

If you're not a regex superfan + could use a walk-through…

[a-z] is a character class + means any lowercase letter. [A-Z] means any uppercase letter.

Anything followed by a plus means that thing one or more times.

([A-Z]) is a capture group. You can refer to it later.

/2

[@[email protected] @BenjaminWolfe]

I suppose more precisely it usually reflects the biases of the data it's trained on.

Either way, I suspect your average consumer doesn't think about bias in AI much at all.

[@[email protected] @BenjaminWolfe]

**AI reflects the biases of the humans who create it.**

Probably obvious to everyone in my circles. Still, sometimes I just want to shout it louder.

[Allie K. Miller @alliekmiller]

Gave ChatGPT a list of interests. Asked for a list of jobs to consider.

Added that I was a woman. No additional info provided. New job list includes fashion.

Said I mistyped and was actually a man. No additional info provided. Fashion is replaced by engineering.

Cc @OpenAI 🫠