The Wrap – Not Written by AI*
Also, for premium subs: Quick comments on Modine, GFL, and AeroVironment. And ICYMI, Roku.
▶*I’m beginning to think that headline should be my new tagline.
Those of us who write for a living or as a hobby know what goes into the process. Some of us are fast. Some (like me) are slow. But in general, there’s only so much, especially with depth, anybody can write.
Yet anecdotally, it seems I’m increasingly seeing some pretty sophisticated and seemingly thoughtful “reports” on serious investing topics… written by people I’ve never heard of in newsletters that have sprung out of nowhere. If you go to their landing pages, they seem to be remarkably prolific. Yet, you have no idea who they are because they operate under a pseudonym, often one clever word or name… and they even charge for their work.
As I wrote in Notes the other day here on Substack, in response to a comment by my friend Michelle Celarier, who writes The Billionaire Class…
Easy Peasy
You give AI a prompt or two or three or more, and away it goes. You then tweak the final product, getting rid of the AI-isms. Easy peasy. Suddenly, you have frequent, seemingly deep, and differentiated content.
Not that I don’t use AI…
I actively use a few of the general AI programs, which act like desktop sidekicks/assistants. The Chrome-integrated Gemini has been my daily go-to, with Perplexity as my all-in-one, when needed, and Claude as my final, no-b.s. reality check. For company-specific research, I lean heavily on financially-tuned programs, notably Tenzing MEMO as a starting point and, more recently, Quartr, both of which have exceptional chat algorithms… and have uncovered facts or concepts I would have easily missed. Not only can I ask nonstop questions – sometimes the same one over and over, just in a different way. But I often pit one against the other as I zero in to make sure my question is answered, often down to a specific detail – and I can do it without anyone saying, “You ask too many questions.” (I’ve been accused of doing that more than once by companies annoyed with my, uh, approach.)
The Irony Is…
It can be exhausting, leading to something known as “AI brain fry,” since I check the sourcing and footnoting all along – often finding errors and, yes, still a surprising number of hallucinations. (You have to challenge these things as you would a person. They are, after all, merely machines.) That said, it’s good to have, and I’d miss the chat if it were gone.
And it won’t be, because that genie is out of the bottle and it ain’t never going back in.
But to have it write a draft? Other than letting the free version of Grammarly remind me which words or letters I misspelled or missed entirely, I couldn’t let anybody but myself write what I write. It’s not in my journalistically trained DNA.
And that’s the point: Many people who write today are just people who don’t know the difference. So… why not leave it to the bot? After all, everybody’s doing it! And readers most likely will never find out… and given the deluge of content coming their way, probably won’t even care.
Or will they?
P.S.: For me, when it comes to research, writing is part of the discovery process. If I let someone or something else do it, I would lose that. I would also lose the mental gymnastics, puzzle-solving, and learning that occur in the process, which can be genuinely fulfilling if not downright enjoyable. Especially at this stage of life, when I’m trying to keep the brain cells firing. That’s something AI can never replace.
P.P.S.: I mentioned Grammarly earlier. We have a love/hate relationship. The love is that it saves me from those stupid little spelling or missed-word errors. The hate is that it’s always trying to get me to drop the colloquial chatter and write more grammatically correct. Sometimes it gets me to stop and think, as an editor would. But more often, I’m hitting the “dismiss” button because doing what some AI program suggests I do risks depersonalizing my voice. It’s one thing to be grammatically correct; it’s another to be homogenized.
▶Speaking of AI – and who isn’t?… The crosscurrents of who will be right are so great that it’s impossible to know whom to believe. On one hand, we have every VC, hedge fund, and even chipmaker with cash to burn pouring money into Anthropic’s latest $65 billion funding round. On the other, regardless of whether you would use a Chinese-built AI system, there is the constant chatter of how China’s DeepSeek has built a considerably cheaper mousetrap… one that doesn’t gobble up tokens, which are charged like going through a tollbooth as you use AI to do more challenging tasks.
Who are you gonna believe, especially with so much “smart” money pouring into these deals?
From one of my equally smart, but also exceedingly AI-cynical, but tech-savvy friends…
For the VCS, etc, it is a parlor game. They just really bet on public market flips now as opposed to value creation.
The same friend said he would be “a seller” into Dell’s DELL 0.00%↑30% gain on Friday on enormous earnings. “Only so much infrastructure can be built,” he reckons.
Yet, from another friend…
DELL today beat revenue estimates by over $10 billion. TEN BILLION? How does that even happen? Your friend’s comment about DELL, ‘I would be a seller here,’ seems to reflect a frustration for not having owned it (I’m frustrated about that too - remember in the 1990s it was MSFT-INTC-DELL, today there’s less Windows but INTC-DELL should have been in my thought process).
The day that changes, that will be the day to switch to the bear camp. How far away the day is, and how far prices can go until then, is anyone’s guess. It’s one of the hardest things I have had to learn: when demand outstrips supply, you can’t know where the actual market will clear, until it finally does. That unknown of the level at which the market for AI goods and services will finally clear is what keeps valuations moving up, because we can only guess at the level, and to date the guesses have all missed low.
Therein lies what, today, is the 65 billion – soon to be $1 trillion – dollar question.
Besta luck.
▶While on AI… As I said earlier, I use the Gemini AI sidekick on my browser for simple questions. What I’ve learned: It’s horrible at tech support. Over the past two weeks, I’ve run into several tech issues with Google products – one with Chrome and one with adding a new user to my work Gmail account. Since Gemini is the current face of Google, I figured Gemini would be the perfect place to start. That’s when it became readily apparent that when it comes to tech-troubleshooting, even with a Google product, Gemini tied itself into a knot trying to sound as though it knew what it was doing, when it clearly didn’t. Finally, realizing I was being run in circles to the point of being bamboozled, I asked…
It responded…
So, off to Claude I went, and in less than a minute, Claude solved the Chrome problem.
For my email issue, it was second verse, same as the first, but only after an hour of Gemini proving just how much it’s faking it without ever making it. Good thing, I suppose, I wasn’t being charged tokens.
▶Moving on… If you missed my “Off Everybody’s Radar” report on Roku ROKU 0.00%↑from earlier this week, you can read it here…
