I think that the post's points on markets not being fully efficient are reasonable, but I also think that any reliable strategy will see its value get eaten pretty quickly, unless it relies on something that only a few people know about.

It seems like the author agrees:

All of the approaches I’ve mentioned are likely to outperform by less than history suggests, due to the increased amount of investment attempting to exploit them. That increased investment over the past decade or two doesn’t look large compared to the magnitude of the inefficiencies, but if trading volume on these etf’s becomes comparable to that of leading etf’s, I’d expect the benefits of these etf’s to go to zero.

The 3% number may already be accounting for this trend, but even if it does, this approach comes with (I assume) a bit more risk than a standard index-fund strategy, plus the need to occasionally rebalance one's portfolio, the fear that comes with sometimes underperforming the market, etc.

And of course, the outside view lets us notice that nearly every person who ever thought they could reliably beat the market was wrong, including many people who were at least as well-informed as the author. That reduces my expectation for the benefits of this strategy. I'd put higher credence on it working that I would for most investor-recommended strategies, since I have a pretty high opinion of the author's past work, but I wouldn't go so far as to advocate that any particular EA follow the strategy to the letter (especially since each person has their own financial strategy/goals).

My approach, as someone who doesn't want to spend a lot of time thinking about small percentage gains/messing around with rebalancing and such, is to just use Betterment (Wealthfront is equally good), an index-fund startup that handles rebalancing and tax benefits for you. I've had solid, slightly-above-market returns from them for years, and the interface is really good.

Bayesian Investor's recommendations are actually pretty similar to the more advanced portfolio recommended in Ben Todd's post Common investing mistakes in the effective altruism community. The article recommends "[adding] tilts to the portfolio for value, momentum and low volatility (either through security selection or asset selection or adding a long-short component) and away from assets owned for noneconomic reasons" as an advanced move that should only be done if "you know what you’re doing."

Likewise, Bayesian Investor's recommended portfolio heavily involves low-volatility and fundamentally weighted (value-tilted) ETFs.

These articles reach fairly similar conclusions because academic research indicates that these strategies have historically outperformed market capitalization–weighted indexes (commonly known as "the market"). Various theories exist about why these strategies outperformed historically and whether they can be expected to outperform in the future.

Your observation that investing is important for EA because it can significantly increase funding for the EA community is why I'm working on Antigravity Investments, a social enterprise with the goal of improving investment returns in the EA community. Right now, we're picking the lowest hanging fruit by recommending that EA organizations move low-interest cash reserves into high-interest and low-risk savings options (see my EA forum article), which is essentially a guaranteed 2.5% improvement in returns every year at current interest rates. If we shift $15 million in cash, that's another $1 million in direct funding for high impact charities over three years.

Interestingly enough our most recommended option is both safer and higher interest than storing large amounts of cash in a checking account.

While there may be obvious things that EAs should be doing, unfortunately it is very difficult to invoke behavior change. My current approach to behavior change with regards to investing is to have an organization with specific expertise in investing help other EAs and EA organizations implement sensible recommendations. This approach seems to be more effective than writing articles online since it removes the prerequisites of having adequate expertise and time to learn about and implement sensible investing practices.

I wrote the high-yielding cash equivalents article because the recommendation seems particularly easy and obvious to implement. So far, although my article was well received, I haven't heard from any EA organization that has attempted to implement the recommendation based on reading the article, although organizations I've directly reached out to in the past have implemented the recommendation. I'm currently in the (very slow) process of doing more direct outreach to EA organizations to determine for them (and for us) whether our recommendation is worth implementing.

To answer your question about whether the advice is worth following, my personal opinion is that some of Bayesian Investor's recommendations are worth diversifying (tilting) into at a level that reflects each investor's confidence about how likely the anomaly will persist into the future. The low volatility factor in particular has achieved very high out-of-sample risk-adjusted and absolute returns, which is promising, but of course a prolonged period of underperformance could be on the horizon—hence the importance of diversifying.