upvote

points

by brotchie3 hours ago |
comments
upvoteby cj3 hours ago|
next
[-]
I love the tax use case.

What scares me though is how I've (still) seen ChatGPT make up numbers in some specific scenarios.

I have a ChatGPT project with all of my bloodwork and a bunch of medical info from the past 10 years uploaded. I think it's more context than ChatGPT can handle at once. When I ask it basic things like "Compare how my lipids have trended over the past 2 years" it will sometimes make up numbers for tests, or it will mix up the dates on a certain data points.

It's usually very small errors that I don't notice until I really study what it's telling me.

And also the opposite problem: A couple days ago I thought I saw an error (when really ChatGPT was right). So I said "No, that number is wrong, find the error" and instead of pushing back and telling me the number was right, it admitted to the error (there was no error) and made up a reason why it was wrong.

Hallucinations have gotten way better compared to a couple years ago, but at least ChatGPT seems to still break down especially when it's overloaded with a ton of context, in my experience.

reply
upvoteby arjie2 hours ago|
parent|
next
[-]
In my case, what I like to do is extract data into machine-readable format and then once the data is appropriately modeled, further actions can use programmatic means to analyze. As an example, I also used Claude Code on my taxes:

1. I keep all my accounts in accounting software (originally Wave, then beancount)

2. Because the machinery is all in programmatically queriable means, the data is not in token-space, only the schema and logic

I then use tax software to prep my professional and personal returns. The LLM acts as a validator, and ensures I've done my accounts right. I have `jmap` pull my mail via IMAP, my Mercury account via a read-only transactions-only token and then I let it compare against my beancount records to make sure I've accounted for things correctly.

For the most part, you want it to be handling very little arithmetic in token-space though the SOTA models can do it pretty flawlessly. I did notice that they would occasionally make arithmetic errors in numerical comparison, but when using them as an assistant you're not using them directly but as a hypothesis generator and a checker tool and if you ask it to write out the reasoning it's pretty damned good.

For me Opus 4.6 in Claude Code was remarkable for this use-case. These days, I just run `,cc accounts` and then look at the newly added accounts in fava and compare with Mercury. This is one of those tedious-to-enter trivial-to-verify use-cases that they excel at.

To be honest, I was fine using Wave, but without machine-access it's software that's dead to me.

reply
upvoteby shepherdjerred3 hours ago|
parent|
prev|
next
[-]
I've gotten better results by telling it "write a Python program to calculate X"
reply
upvoteby brotchie1 hours ago|
parent|
next
[-]
For the tax thing. I had Claude write a CLI and a prompt for Gemini Flash 2.5 to do the structured extraction: i.e. .pdf -> JSON. The JSON schema was pretty flexible, and open to interpretation by Gemini, so it didn't produce 100% consistent JSON structures.

To then "aggregate" all of the json outputs, I had Claude look at the json outputs, and then iterate on a Python tool to programmatically do it. I saw it iterating a few times on this: write the most naive Python tool, run it, throws exception, rinse and repeat, until it was able to parse all the json files sensibly.

reply
upvoteby dmd2 hours ago|
parent|
prev|
next
[-]
Yeah, in my user prompt I have "Whenever you are asked to perform any operation which could be done deterministically by a program, you should write a program to do it that way and feed it the data, rather than thinking through the problem on your own." It's worked wonders.
reply
upvoteby 2 hours ago|
parent|
[-]
deleted
reply
upvoteby cj2 hours ago|
parent|
prev|
[-]
Good call. I’ve also had better results pre-processing PDFs, extracting data into structured format, and then running prompts against that.

Which should pair well with the “write a script” tactic.

reply
upvoteby tavavex2 hours ago|
parent|
[-]
Yeah, asking for a tool to do a thing is almost always better than asking for the thing directly, I find. LLMs are kind of not there in terms of always being correct with large batches of data. And when you ask for a script, you can actually verify what's going on in there, without taking leaps of faith.
reply
upvoteby ElFitz2 hours ago|
parent|
prev|
[-]
I’d say for these use cases it’s better to make it build the tools that do the thing than to make it doing the thing itself.

And it usually takes just as long.

reply
upvoteby thijsvandien3 hours ago|
prev|
next
[-]
I don't know, but I would never upload such sensitive information to a service like that (local models FTW!) or trust the numbers.
reply
upvoteby basch3 hours ago|
parent|
[-]
Which part is sensitive? Social is public, income is private but what is someone going to do with it?
reply
upvoteby jumpman5002 hours ago|
parent|
next
[-]
It's not good in some job negotiations if someone has a very clear picture of what your current net worth and income is. Also in some purchases companies could price discriminate more effectively against you.
reply
upvoteby thijsvandien2 hours ago|
parent|
prev|
[-]
Now that's a question I'd feel more confident having answered by an LLM. Personally, I'm tired of arguing with "nothing to hide", which (no offense) is just terribly naive these days.
reply
upvoteby mandeepj2 hours ago|
prev|
next
[-]
> Result was amazing. I now actually fully understand the tax position.

You couldn’t do that with TurboTax or block’s tax file? You don’t have to submit or pay.

reply
upvoteby MikeNotThePope2 hours ago|
prev|
next
[-]
Be careful with taxes. Hallucinations will cost you.
reply
upvoteby whattheheckheck3 hours ago|
prev|
next
[-]
I had ai hallucinate that you can use different container images at runtime for emr serverless. That was incorrect its only at application creation time.

Hope you dont get audited

reply
upvoteby slopinthebag2 hours ago|
prev|
[-]
> had Claude prepare all of my docs for upload to my accountant: FinCEN reporting, summary of all docs, etc.

I imagine your accountant had the same reaction I do when an amateur shows me their vibe codebase.

reply