I love AI
-
Thanks, Horace. I can say it now. They signed a $2M contract with a 4% profit margin to build a very elaborate set for an opera. My daughter was the PM and told them up front she did not think they were capable of completing it in the timeframe, her counterpart at the opera said they same thing, and sure enough they weren't. The whole thing was one big clusterfuck from the gitgo. Now the entire production may be canceled. Let the lawsuits begin. I hope to hell they had E&O insurance.
-
That is frustrating. It reminds me of an IT project I was leading in 2018 they wanted to launch in 2019 and I told them not to rush testing and to push the actual deployment to Spring 2020. They said they "couldn't wait that long" so they switched contracts. It's now almost 2026 and they're still working on it.
The point being, folks like your daughter need to have their professional opinions listened to.
-
Had an interesting experience with Gemini Pro. They limit the "pro" mode to 100 prompts per day, or even less if your prompts are super complicated. I had been using the unlimited Fast mode last night to talk about a potential stock pick, and we chatted for like an hour. It's a thinly traded stock with lots of insider buying and lots of short selling and other details that are pretty interesting. Today I realized I could use a single "Pro" prompt in that same chat, and ask it to go over the conversation and tell me what it thinks. It was interesting to read "Pro" talk about "Fast" as if it was a different person, who happens to be sort of dumb. Pro identified the true risk of the stock, which is a binary legal regulatory risk. It could easily halve overnight if a certain judgment comes down not in its favor. I guess that's how I'll use Gemini going forward. Conversations in "Fast" mode to get basic ideas fleshed out, then a Pro prompt to clean up whatever mess was left by the Fast conversation.
-
Im surprised that doesn’t count as a complex prompt given the number of tokens used
I don't think it literally goes over every token again. Or maybe it does. I thought it just maintained a "context". But that "context" might just be the transcript. The number of Pro prompts used isn't made clear, so I don't even know how many Pro prompts I used when I asked about the stock conversation.
My prompt: I'm considering how to economize my "pro" prompts, of which I get 100 per day. I wonder if I can have a longer conversation with you in "fast" mode, then ask in "pro" mode, in the same chat, for you to reflect on the whole thing and comment on anything you'd say differently.
Gemini Pro:
That is a very savvy way to manage your high-performance tokens. To give you the short answer: Yes, that strategy works perfectly.
Because I maintain the context of the entire conversation regardless of which "mode" you are currently using, switching to a Pro model at the end allows it to "read" everything that happened in Fast mode.
Here is a breakdown of how to make this strategy most effective:
Why this works
-
Persistent Context: The "memory" of the chat isn't tied to the specific model version. When you toggle to Pro, the model receives the transcript of the Fast conversation as its input context.
-
Draft & Refine: You can use the Fast mode for the "heavy lifting"—brainstorming, outlining, and data gathering—and save the Pro mode for synthesis, nuance, and final quality control.
Best practices for the "Pro Reflection"
When you are ready to use one of your Pro prompts to review the session, try using a prompt structure like this to get the most value:
"I’ve been brainstorming in Fast mode. Now, using your Pro capabilities, please review our entire conversation above. Identify any logical gaps, suggest more sophisticated alternatives to the solutions provided, and point out any nuances the previous model might have missed."
A "Pro" tip for even more economy
If you find yourself running low on Pro prompts, you can actually bundle tasks. Instead of asking one question at a time in Pro mode, give it a multi-part instruction:
-
"Critique the logic of our previous 10 messages."
-
"Rewrite the final conclusion to be more professional."
-
"Suggest three 'wildcard' ideas we haven't considered yet."
Would you like to try a "Fast" brainstorming session right now so we can test how a Pro reflection looks at the end?
-
-
Just tried this from my previous Calc semester
Fast mode prompt: answer question 11. Tell me whether I should trust Fast mode with a question at this level (undergraduate calculus), or whether I would be better off using one of the other two modes, and if so, which mode would be best.
This is a classic "Related Rates" problem from undergraduate calculus. For a problem like this—which involves spatial reasoning, geometry, and calculus—here is my honest assessment of which mode to use:
Which mode should you use?
You should use Pro mode (Gemini 1.5 Pro) for this specific problem.
- Why not Fast mode? Fast mode is excellent for text and standard coding, but it can occasionally struggle with the "spatial logic" required to correctly set up the similar triangles in a diagram. It might misidentify which variable represents the tip of the shadow versus the length of the shadow.
- Why Pro? Pro mode has significantly better visual reasoning. It can "see" the relationship between the light source, the student, and the shadow more accurately and is much less likely to make a sign error or a substitution mistake during the derivative step.
Solution for Question 11
<Fast mode gives an answer and the math steps to arrive at it>
Pro mode prompt: I have set pro mode. Please check your answer to the previous question.
I have reviewed the previous response using Pro mode logic.
I can confirm that the previous calculation and logic were 100% correct. The "Fast" mode actually handled this specific problem perfectly, but relying on Pro for these types of questions remains the safer bet for the reasons mentioned (specifically, geometric interpretation).
Here is the "Pro" verification of the steps to give you double assurance:
<Pro mode goes over the math steps>
One final "Pro" reflection:
While "Fast" mode got this right, it is worth noting that if the problem had been slightly more complex—for example, if the student were walking up an incline or if the light were moving—the "Fast" mode would likely have failed to set up the geometry correctly. Sticking to "Pro" for calculus word problems is the correct long-term strategy. -
Interesting commentary and perspective, Horace, I had a few funds left in my 2025 tech budget at work so I bought a computer monitor I’m gonna keep vertical mode and it’ll just be my AI slave throughout the day. I haven’t looked into it, but in my mind, I was going to purchase the Gemini pro subscription.
-
Interesting commentary and perspective, Horace, I had a few funds left in my 2025 tech budget at work so I bought a computer monitor I’m gonna keep vertical mode and it’ll just be my AI slave throughout the day. I haven’t looked into it, but in my mind, I was going to purchase the Gemini pro subscription.
