Live data from Hacker News

AI financial advice is surprisingly good, especially if you ask right questions

mitsloan.mit.edu

411–420 of 443 posts

Re: AI financial advice is surprisingly good, especially if you ask right questions

#412
post #28
post #5

I use YNAB ( https://www.ynab.com/ ) for budgeting so I already had all of my financial data in a single source. Exporting the CSVs locally and asking Claude to be my financial advisor legitimately gave me good advice. Not just nagging me to save more (which is always useful), but how to organize my budget categories better, detecting longer term spending patterns I wasn't thinking much about, researching credit card…

>YNAB I've fallen behind on keeping it up-to-date, but I feel I ought to plug anything self-hostable: https://actualbudget.org/

This is the big pain-point for me with self-hosted budgeting software.

The corporate players in the space can jump through the security hoops and integrate with Plaid, but if I wanted to automatically pull my own data into the single-user budgeting tool I host entirely on my own network I need to pass the same security checks as a full-on fintech startup.

Currently I enter transactions manually, but manually reviewing multiple credit card and bank statements is a chore and I don't keep up with it as much as I should. To make things worse, since I ideally want a live view of my budget and not a post-statement-issuance monthly review, I'm constantly checking over transactions I've already entered when their ordering changes in my credit card company's app.

The closest thing to a potential solution involves extracting my Monarch Money session key from my browser and pulling data from their GraphQL endpoints with the help of an unofficial Python SDK. That doesn't solve the "relying on a subscription service" problem, but it would at least let me get my data into the custom budgeting tool I built that actually meets my needs.

Re: AI financial advice is surprisingly good, especially if you ask right questions

#413

Earlier quoted context omitted.

Your model/harness will indiscriminately do web searches to get answers. I believe that’s where the real risk is.

I don't know for sure, but will it require something like with high "SEO" rankings to fake the reputation of the sources? So if the LLM search for a specific spreaded knowledge splitted across many bad sites, it can poison the model maybe.

You mean, how difficult will it be for an attacker to get their text seen by the LLM that does arbitrary searches?

It could be as simple as a malicious prospectus for AcmeCo, and then try to get AcmeCo on the radar so that the LLM-tools find your document and incorporate it. The malicious bits don't even need to be AcmeCo-related, they could be to pump (or dump) practically anything.

Re: AI financial advice is surprisingly good, especially if you ask right questions

#414
post #5

I use YNAB ( https://www.ynab.com/ ) for budgeting so I already had all of my financial data in a single source. Exporting the CSVs locally and asking Claude to be my financial advisor legitimately gave me good advice. Not just nagging me to save more (which is always useful), but how to organize my budget categories better, detecting longer term spending patterns I wasn't thinking much about, researching credit card…

> better adapt quickly

maybe they can grow larger brains and an extra hand so they can hold 3 calculators at once?

Re: AI financial advice is surprisingly good, especially if you ask right questions

#415
post #28

Earlier quoted context omitted.

>YNAB I've fallen behind on keeping it up-to-date, but I feel I ought to plug anything self-hostable: https://actualbudget.org/

This is the big pain-point for me with self-hosted budgeting software. The corporate players in the space can jump through the security hoops and integrate with Plaid, but if I wanted to automatically pull my own data into the single-user budgeting tool I host entirely on my own network I need to pass the same security checks as a full-on fintech startup. Currently I enter transactions manually, but manually reviewin…

If I had my 'druthers there would be legal rules around the rights of customers to access their information, using an API key limited to reading past/pending transaction data.

I suppose an extra-paranoia feature might be letting people set a salt value for any data which is too sensitive to release but it needs to be present for comparison/sync purposes.

Re: AI financial advice is surprisingly good, especially if you ask right questions

#416
post #398
post #168

Earlier quoted context omitted.

Most people here are probably paid too much to contribute to a Roth IRA.

But you can do a backdoor Roth IRA by maxing out your 401k and then rolling it over into a Roth IRA with recharachterization. You lose the tax benefits of the 401k but gain the tax benefits of the Roth IRA which can be a good trade depending on your tax situation.

No, this isn't how a backdoor Roth IRA works.

Backdoor Roth IRAs involve making a traditional IRA contribution and not taking the deduction at tax filing time (because you can't), but then rolling over (not recharacterizing, that's something else) the traditional IRA contribution into a Roth IRA. It's completely tax free, assuming you have a $0 traditional IRA balance once the rollover is complete. The usual way to accomplish this is to roll all traditional deductible IRA balances into a traditional 401k first.

What you're kind of thinking of, but also not quite right, is called a 'mega backdoor Roth', which involves contributing to a 401k via a non-deductible contribution (which is not part of 24.5k/yr limit), then immediately rolling it over into a Roth 401k. It has to be allowed by the plan, but some plans even offer to do the rollovers for you automatically.

The Mega Backdoor basically lets you get an extra ~40k/yr of Roth contributions, if you can afford it.

Re: AI financial advice is surprisingly good, especially if you ask right questions

#417

Earlier quoted context omitted.

I believe the general idea is to max it if you can. If you can't, put whatever you can, and forego luxuries like vacations until you can.

Right, so you max your 401k instead of diversifying a portion into something more liquid, and when something happens before 55 you have to take a major bite out of it just to access that money?

That would be what an emergency fund is for, in traditional financial literature. If you mean something more tragic, that's what insurance and other products are for. Even then, 401ks have exceptions for something like total disability that allows early withdrawals without penalty.

If you just mean "I want my money early", well... there's lots of ways to access retirement account money early without penalty with some planning.

Re: AI financial advice is surprisingly good, especially if you ask right questions

#418
post #283

Earlier quoted context omitted.

> The seller explained us how a lease was so much better financially than outright buying. Of course a lease is better than buying.. for the dealer. So the seller was honest, just not entirely transparent. Leasing is one of the dumbest financial moves ever. Forces you into a perpetual treadmill of payments for life. Just buy a car, pay it off (cash if you can, or with a loan) and then drive it forever.

Some people want a new car all the time. Leasing is not a bad way to do that. Wanting a new car all the time is the expensive decision, not leasing, which is merely a manifestation of that decision. As someone who drives a ten-year-old car that probably has several good years left, I completely understand why someone would not want my boring life. This is the fallacy of people who recount stories of the “millionaire…

> Some people want a new car all the time. Leasing is not a bad way to do that.

True.

But does that make any economic sense? If you're so rich that it doesn't matter, sure why not.

But for most people leasing is a terrible economic decision. If money matters, never lease. If your wealth is past the point where money matters, go for it.

> As someone who drives a ten-year-old car that probably has several good years left

I'd say several decades at least. My new car is 13 years old, which I consider barely broken in.

Second oldest is 22 years old, doing great. My other cars are much older, in their third and fourth decades.

Barring a crash, older cars can go on for a very long time. Newer cars built these days have a very limited lifetime ahead of them, as the electronics will be unrepairable in short order.

Re: AI financial advice is surprisingly good, especially if you ask right questions

#420
post #296

Earlier quoted context omitted.

You are suffering from the survivorship bias in its purest form. I really hope nobody follows your advice.

And experiences come from the bulliest of bull markets. Where most people who bet on large sector are winners. Past performance is most likely future performance if things go south...

> And experiences come from the bulliest of bull markets.

That is very true and a trap that is important to be aware of. Anyone who started investing after ~2010 has only ever experience a bull market, so expect severe pain at some point. It doesn't go up forever.

I'm old enough that my investing history inclused the dot.com implosion that wiped out nearly all of everything, and also the financial crisis of 2008. It remains true that investing in the winning companies is by far the best return by orders of magnitude.

Post reply on HN