Even a broken clock is right twice a day
Open your feed on any given night and you can find three ways to get rich before the video ends. A physician wipes out her W-2 tax bill with an Airbnb. A 26-year-old has an AI model build a portfolio that beats the S&P 500. Someone explains the loophole the wealthy don't want you to know about.
Most advisors respond by calling all of it dangerous nonsense. I understand the instinct. I don't think it helps you.
Here is the problem with that reaction: a lot of what these accounts describe is real. The tax provisions exist. The strategies work under the right facts. Somebody, somewhere, ran the play and it landed exactly the way the video said it would. When an advisor calls a genuine provision of the tax code a scam, the viewer knows better, keeps scrolling, and stops listening to advisors altogether.
So I want to be more careful than that. Not to be diplomatic, but because the imprecision is what gets people hurt.
A broken clock is right twice a day
These strategies are rarely invented out of nothing. They usually trace back to a real moment when the thing genuinely worked. A specific year, a specific tax code, a specific set of facts, a specific person who met every condition. That origin is real. It is why the story has legs.
The trouble is that the clock stopped. The conditions that made it work were part of the result, and they don't travel with the video. Twice a day the stopped clock lines up with the real time and looks authoritative. The rest of the day it is confidently wrong, and it never tells you which moment you are in.
That is the honest version of the warning. Not "this is fake." Rather: this was true somewhere, under conditions you may not share, and the clip won't tell you the difference.
The deduction is real. So are the conditions.
The clearest example right now is accelerated depreciation on rental property.
The One Big Beautiful Bill Act, signed July 4, 2025, permanently restored 100% bonus depreciation for qualifying property acquired after January 19, 2025. Pair it with a cost segregation study and a high earner can produce a six-figure first-year loss on a rental. That is not a trick. It is the tax code working the way Congress wrote it.
The condition is the part that gets cut for time. Rental losses are passive by default and cannot offset a salary. There are two ways around that. Real estate professional status requires more than 750 hours a year and more than half of your working time in real estate, which rules out nearly every full-time W-2 professional. The short-term rental route is the one most of these videos are actually selling, and it is legitimate. If the average guest stay is seven days or less, professional status is not required. You still have to materially participate, generally 100 or more hours and more than anyone else, your cleaner and your property manager included, or 500 hours total, backed by contemporaneous logs, in the year the property goes into service.
Read that again, because it is the whole ballgame. Contemporaneous logs. Hours you actually worked. More than the person you hired.
The return that already came back
Here is the part that concerns me most, and the reason I wanted to write this at all.
A tax return that was accepted is not a tax return that was approved. The IRS does not review most returns when they are filed. The deduction went through, the refund landed, the strategy appeared to work, and you have every reason to believe you did it correctly. That feeling of confirmation is powerful, and it is not evidence.
The assessment window generally runs three years, and longer when income is substantially understated. Material participation cases are decided on documentation, and taxpayers lose them regularly for the same reason: the hours were real, but the records were reconstructed afterward. Meanwhile depreciation is a deferral, not a forgiveness. When you sell, recapture brings it back, and the components reclassified in the cost segregation study come back at ordinary rates.
So the person who followed the video may not find out for years. When they do, it arrives as back tax, interest, and penalties, usually in a year they had already planned around money they thought was theirs. That is a genuinely rough thing to watch happen to someone, and it is almost always avoidable.
The backtest already knew the ending
The AI portfolio videos have their own version of this.
When someone asks a language model to pick stocks "as of 2019" and then shows how the picks did, the model was trained on what happened after 2019. Researchers call this look-ahead bias. A 2025 study that had GPT-4 pick stocks every year back to 1985 found significant positive alpha in only about one year in four, and the authors flagged the model's later training data as a limitation on even that result.
A chart built by a model that already knows the answer is a picture of the past with a new caption. The clock is right, but only because someone set it to match.
What the data says about the messengers
It is worth being fair to the people posting, too.
Ali Kakhbod and his co-authors studied recommendations from more than 29,000 finfluencers on StockTwits. About 28% were genuinely skilled, generating 2.6% monthly abnormal returns. The majority, 56%, had negative skill: following them cost roughly 2.3% a month. The uncomfortable finding is the distribution. The negative-skill group had more followers and more influence on retail trading, and the most frequent posters were the least likely to be skilled.
So the honest statement is narrower than "influencers are wrong." Roughly a quarter of them know exactly what they are talking about. The algorithm does not show you those first, because certainty performs better than accuracy.
How to tell what time it is
None of this means you should ignore what you see. It means the video is the beginning of the work, not the end of it. Whether a strategy comes from a feed or from an advisor, these are the questions that separate a real opportunity from a stopped clock:
- Who is this built for, and do I meet every condition, or just the headline one?
- What has to stay true for this to keep working, and what happens if it changes?
- What does the exit look like: the sale, the recapture, the unwind, the margin call?
- If this is examined in three years, what do I hand over?
- How many people tried this and did not post about it?
- Does the person telling me get paid when I act on it?
The last question applies to me too. Ask it.
Bring it in before you build on it
If you saw something that looked compelling, I would rather you bring it to us than talk yourself out of it quietly. Sometimes the answer is that it is a real strategy and we should build it properly, with the documentation and the exit planned from day one. Sometimes the answer is that it was right for someone else, in a different year, under facts you do not have. Both are useful answers, and the second one is far cheaper to receive now than in an examination letter.
That is most of what experience buys in this business. Not knowing which strategies exist. Knowing how often each one actually works, for whom, and what it costs the person who arrives second.
Sources (internal reference for compliance review)
• Bonus depreciation and passive activity rules: Tax Strategy Playbook; Angel Investors Network; Guest Manual. Verify against IRC §469, §168(k), Treas. Reg. §1.469-5T, and IRS Notice 2026-11 before publishing.
• Assessment window and material participation substantiation: verify against IRC §6501 and Treas. Reg. §1.469-5T(f)(4) before publishing.
• "Could ChatGPT have earned abnormal returns?" Modern Finance (2025): mf-journal.com
• Kakhbod, Kazempour, Livdan, Schuerhoff, "Finfluencers": SSRN