Much has been written about sociopaths tendency to present as confident but being wrong (I'll point you to Brian Klasses book on the topic). But AI is especially good at this:
This morning I asked Claude to answer a question (to save me time, as I didn't want to do the reading myself).
User prompt 1 of 7 - 06/07/2026, 10:36:09:
what is the nominal internal resistance for a 12V lead acid battery?
Claude Haiku 4.5:
The nominal internal resistance of a 12V lead acid battery is typically between 5 and 15 milliohms (mΩ), depending on the battery's size, design, and condition.
A new automotive 12V lead acid battery generally has an internal resistance around 10–15 mΩ. However, this varies based on factors like:
- Battery capacity: Larger batteries (with more plates and electrolyte volume) typically have lower internal resistance.
- Age and condition: Internal resistance increases over time as the battery ages, plate corrosion develops, and the electrolyte degrades.
- Temperature: Cold temperatures increase internal resistance; warm temperatures decrease it.
- Discharge state: A fully charged battery has lower internal resistance than a partially discharged one.
For reference, this internal resistance is why lead acid batteries can deliver high discharge currents (important for starting engines) but also means they have significant voltage drop under heavy load.
Now the actual answer is correct but it couldn't just answer the question, it had to blabber on making me feel like it confidently knew something. It didn't ... (pretty much like a politician).
Underline mine. I noted a couple of things (and I also know a fair amount about batteries and chemistry) and because of this un-needed point about internal resistance I smelled something rotten in the state of AI, and so I began strategically with:
User prompt 2 of 7 :
interesting:
> For reference, this internal resistance is why lead acid batteries can deliver high discharge currents (important for starting engines) but also means they have significant voltage drop under heavy load.
could you compare and contrast this with LiFePO4 please?
Claude Haiku 4.5:
LiFePO4 vs Lead Acid Battery Characteristics
...
blithering omitted for brevity, but ends it with:
...
Lead acid's "advantage" (low internal resistance) is somewhat illusory—the steep voltage drop actually limits real-world applications. You need the battery to be relatively large relative to your load.
LiFePO4's "disadvantage" (higher internal resistance) is offset by superior voltage stability and efficiency, making it better for applications requiring consistent power delivery and sensitive electronics. LiFePO4 also doesn't degrade as quickly from deep discharges.
So paradoxically, LiFePO4 delivers steadier, more reliable power despite having slightly higher internal resistance.
Interesting ...
User prompt 3 of 7 ...:
so in light of this above statement can you expand on what you mean back here:
For reference, this internal resistance is why lead acid batteries can deliver high discharge currents (important for starting engines)
This is where I found it interesting to read what claude was thining, because between the ClaudeHaiku 4.5 prompt and the "you're right to push back" we see some thinking. I'll insert that below as it appeared in the transcript (underline mine).
Claude Haiku 4.5:
The user is asking me to expand on my earlier statement about internal resistance being the reason lead acid batteries can deliver high discharge currents. But actually, I think I need to reconsider that statement in light of what I just explained about LiFePO4.
Looking back at what I said:
"this internal resistance is why lead acid batteries can deliver high discharge currents"
This statement is actually misleading or incorrect. Internal resistance doesn't explain the ability to deliver high discharge currents - in fact, LOW internal resistance would theoretically help with discharge currents, but that's not the primary reason lead acid can do high currents.
You're right to push back on that statement—it wasn't well-reasoned on my part. Let me clarify.
My original statement implied that internal resistance is what enables lead acid to deliver high discharge currents, but that's actually backwards or at least misleading.
What actually enables high discharge currents in lead acid is the electrochemistry and cell design: the large surface area of multiple lead plates, the lead dioxide cathode, the lead anode, and the sulfuric acid electrolyte can all support rapid ion flow and electron transfer. The internal resistance of 10–15 m? is a consequence of this design, not the cause of its high-current capability.
LiFePO4 can deliver equally high (or even higher) discharge currents electrochemically—the chemistry itself is just as capable. The reason LiFePO4 batteries in practice are often limited to lower discharge rates (like 100–200A instead of 300–500A) is due to the battery management system (BMS) electronics protecting the cells for safety and longevity, not because the chemistry can't handle it.
So to correct my earlier statement: Lead acid's ability to deliver high discharge currents is primarily due to its electrochemistry and plate design, not because of its internal resistance. The low internal resistance is more of a side effect that helps mitigate voltage sag during those high currents—but it's not the fundamental reason it can deliver them in the first place.
Thank you for catching that imprecision.
So why is this here? Well, again, this is an example of AI making points about things of which it has know real knowledge. Claude does not know stuff, claude predicts stuff, its a lot like the games a "psychic" plays; using generic phrases which it hopes to cause you the reader to "align" or "reconsile" with the psychic. Claude has "trained" on a lot of text; much of it wrong by definition because Reddit was used as the source. Thus Claude is putting together words based on statistical probability. Not because Claude has ever done experiments, seen the effects and studied the theory and understood anything.
Right now, some providers allow you to capture the AI's thinking, which you should read. You have to download the transcript however. I'll be sad when this (inevitably) goes away.
My point is this:
Tools like AI are a very helpful reference if you already know something about a topic, but be very wary of using them as learning tools, if you've ever been on Reddit you'll know that so many kiddiez there will say stuff that is plain wrong because they don't know any better.
AI was trained on a poisioned chalice and it of course doesn't know.
Be careful out there
.jpg)
