AI - confidently wrong

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

Watch Carefully

I watched Ford vs Ferrari the other night and loved it. However right at the end I went "wait on a minute" hit pause and had to go back because suddenly Shelby was wearing a different watch. So I watched the end of the movie but the next morning I wanted to explore what I'd seen. So I grabbed a few shots and tweaked the gamma and found this:

I wish there was better to be had, but it is not any watch Shelby was wearing in the rest of the movie, and nor is it Ken Mile's watch.

I posted on Reddit a question but got nothing (well stupidity) and one suggestion I'll come back to. I asked the Google prompt (AI) (which is based on the majority opinion of whatever any twaddle is written on Reddit) has only the two well known versions.

Now we know its neither of those because the Carerra is a bicompax design (and this is tricompax) and we know that Kens watch hasa black face with white subdials (seen here clearly while he was doing pre-race reviewing of the track).

getting back to that Redditor suggestion; he proposed that it may be a Rolex Daytona ...

which it might well be, as from this angle the black bezel leans in that direction. If I try hard enough I can imagine the Rolex Crown on the watch crown.

What's so strange to me is that nobody else has noticed this little mystery.

Blew a fuse

So, the other day I went for a ride and found that my headlight wasn't working (I was on a short ride). I tried high beam and it was, so I thought "blown bulb" however ...  it was then I noticed that my neutral light was very dim and that the indicators were not working properly (not blinking).

After I got home I pulled the headlight off and checked the bulb.

All was fine, which wasn't a relief, because now I had to find what the problem really was. I lifted the seat and found that a fuse (was blown). 

Now its worth revisiting the bike and note that its not "original". As you can see below the airbox is missing and you can make out a tray under the seat which holds most of the electricals:

Including the slim battery:

which (importantly) has a "quick plug" type connection with a built in main fuse (of the blade type).

It was rated at 15A and I noted (in the manual) that the main fuse should be a 20A. I replaced the fuse then I checked wires to look for some sort of short (wire rubbing against inside of the metal bucket is a candidate) and all appears fine.

This leaves me with 2 questions:

1. what caused the fuse to blow
2. how then (with the fuse being blown) was my high beam getting power?

So I pulled the fuse and put my Multimeter (set to Amps and it reads up to 10) in across its place to see when current was flowing and began looking for problems. I quickly found that the headlight worked sometimes, but with a bit of movement went off and the meter rapidly climbed. I'd then turn off the light and resume after "moving it till it was ok again".

It took me a couple of days (no more than 4 hours each session) to work out that it was actually the loom rubbing against the steel of the bike ... look carefully at the picture below, you can see the loom resting against the steel of the headlight bracket.

I became confident this was the issue because when I relieved this pressure / contact (lifting up) the much higher power drain went away and the headlight operated normally.

I looked around and found some things worth observing:

1. weight pulling down on the harness from the cable

The loom ran under the cable for the exhaust valve decomression cable and

2.

The loom ran under a bracket which supports the tank from wobble. I thought perhaps if this was lifted up it would free the loom from fouling the bracket and consulted the manual:

ahhh hah ... so somehow this had been put together wrong some time back (probably 2016 when it was imported into Australia).

So I re-ran the cable back under the loom and ran the cable over the support:

which then lifted the loom away from that cable seen here:

which you can see elevated it and also took the weight off by running that cable for the exhaust decompression to under where the loom goes , this now relieved the headlight support bracket rubbing point.

3. I also "splinted" with some plastic (cut a section from a 2L milk bottle), some duct tape and a cable tie (which should also prevent it pulling out of that if it feels so inclined).

So, its now been test ridden and seems to be doing ok ... 

I fully appreciate that there will be some copper getting atmosphere in there now, which is less than ideal, but to do that job I'd have to strip back even more than I did as well as undo the loom searching for the wire, cutting and joining in a new piece and putting it all back together (probably adding several hours).

Lets see how long it lasts :-)

Claude the confident bullshitter

Its really interesting now that DuckAI is providing the "thinking" responce of its AI's as well as its output. So I asked Claude:

User prompt 1 of 2 - 5/25/2026, 3:19:41 PM:
what is the etymology of the finnish word 'myrsky'

and it thought and answered:

"Myrsky" (meaning "storm" in Finnish) has roots in Old Norse. The word is believed to derive from Old Norse *mórskaðr* or related forms, which refer to storms or rough seas. It's part of a broader Germanic/Scandinavian word family connected to words meaning storm, tempest, or violent weather phenomena.

The word entered Finnish through contact with Scandinavian languages, particularly during the medieval period when Swedish and other Nordic languages had significant influence on Finnish through trade, cultural exchange, and Swedish rule in parts of Finland. This is a pattern seen with many Finnish words that have Scandinavian origins.

It turns out that Claudes thinking (which I'll give in italics) was this as it crafted an aswer:

Claude Haiku 4.5:
The user is asking about the etymology of the Finnish word "myrsky". This is a straightforward linguistic/etymological question about a specific word...

Basically thinking "That's easy!!" ... Well, the answer sounded ok  to me.  So (being interested) I asked a follow up question

User prompt 2 of 2 - 5/25/2026, 3:20:29 PM:
so, in Old Norse does it mean dark sky?

It was here I noticed a lot more activity on the thinking. So, when I reviewed  the thinking, despite its confidence it turns out that Claude may not have been so certain:

Claude Haiku 4.5:
Actually, I'm not entirely certain about the exact Old Norse etymology ...

The thinking as it turns out was:

The user is asking whether in Old Norse the word means "dark sky." I need to be careful here because I made a claim about the Old Norse etymology in my previous response, but I should verify if that's accurate and whether it specifically relates to "dark sky."...

right ... 

So, AI is not even "second hand thinking" its "hold my beer while I go dunning kruger" in this case (and perhaps many cases).

The scene if you don't know it:

Just so you know ...