And not working with anyone else.

AI written code is often much easier to read/review than some of my coworkers'

Can't say the same for my colleagues' AI written code. It's overtly verbose and always does more that what's required.

As a sister comment said, floating point computations are commutative, but not associative.

a * b = b * a for all "normal" floating point numbers.

Yup. This is the one.

The Trilby series is fantastic!

I think what he is referring to is this:

If one battery is 50%, and the other at 70%, and you put both in, one will end up at 80% and the other at 100%. When one is full, those cheap chargers stop charging the pair.

That's not how it works at all.

When one is 80% and the other is 100%, the full one enters a state of over charge, reaching 105% or even 110% charge. This is safe.

The H2 reaction then rapidly speeds up, leaking energy in the form of heat. The full battery heats up from overcharge but is otherwise safe.

You end up with both batteries at 100% and maybe 110%, and a day or two later the 110% overcharge settles down to 100% by leaking out.

-------

So you waste a bit of power but as long as the trickle charge is safe and as long as the overcharge is only for a few dozen hours or so, it's fine. In the very long term (if you keep doing this) the NiMH could get damaged. But if we are talking about a once-per-yeqr top off charge, then it's fine.

The problem is like I said before: the safe rate of overcharge is low. This means that these chargers must charge slowly, maybe 10 hours or longer.

Any faster risks blowing through the NiMH innate ability to take an overcharge and convert it into heat. (This results in a forceful vent, a 'pop' sound that permanently damages the NiMH as the H2 gas escapes the safety hatch).

------

Note that these super cheap chargers are simply a glorified 10+ hour timer. They don't even check the state of the AA cells.

So if you stick a 50% full battery in, it will charge the battery to 150%, most likely. (But safely, as the NiMH just leaks out the excess energy as heat, as I said earlier).

Ok, but do you agree that in this day and age of cheap electronics, it is really not worth it to even consider all this and it is better to just make a charger that doesn't overcharge and doesn't come with extra instructions?

It doesn't really make sense to buy a $20 charger when you only have $16 or so worth of AA cells (ie: a 16 pack from Amazon Basics). The $10 charger makes a lot more sense.

As for the knowledge problem: that's easily solved. I already told you what to do so you know how to work the $10 charger now. If you still want the $20 charger that's fine, it's your money.

I suppose there are lots of different chargers. The one I had made both charging lights turn off when one turned green. Maybe it was overcharging but still showing as red till both were charged?

I've been using NiMH AA batteries for over 20 years. I've never had one leak. It's not unique to Eneloop.

> those controllers go forever between charges.

In my experience - about a month. Less if you play more often.

Eneloops were huge for use in DSLRs.

And they are overhyped.

Most of them are about 2000 mAh. Other NiMH batteries can have, say, 2700 mAh. So even though the latter have a higher discharge rate - after 6 months of storage the latter still has more juice.

The benefit with the 2700 mAh, of course, is if you're using when full, you can use it for much longer.

If they cost the same, I could see the hype. But most people are still better off with regular NiMH AA batteries.

Eneloop Pro cells have a rated capacity of 2500mAh.

I can't think of any good applications for conventional NiMH cells any more - they're dominated by LSD NiMH cells in low-discharge applications, by lithium primary cells in ultra-low-discharge applications and by the various lithium secondary chemistries in high-discharge applications.

Just curious - how does it compare to GLM 4.7? Ever since they gave the $28/year deal, I've been using it for personal projects and am very happy with it (via opencode).

https://z.ai/subscribe

There's no comparison. GLM 4.7 is fine and reasonably competent at writing code, but K2.5 is right up there with something like Sonnet 4.5. it's the first time I can use an open-source model and not immediately tell the difference between it and top-end models from Anthropic and OpenAI.

Kimi k2.5 is a beast, speaks very human like (k2 was also good at this) and completes whatever I throw at it. However, the glm quarterly coding plan is too good of a deal. The Christmas deal ends today, so I’d still suggest to stick to it. There will always come a better model.

From what people say, it's better than GLM 4.7 (and I guess DeepSeek 3.2)

But it's also like... 10x the price per output token on any of the providers I've looked at.

I don't feel it's 10x the value. It's still much cheaper than paying by the token for Sonnet or Opus, but if you have a subscribed plan from the Big 3 (OpenAI, Anthropic, Google) it's much better value for $$.

Comes down to ethical or openness reasons to use it I guess.

Exactly. For the price it has to beat Claude and GPT, unless you have budget for both. I just let GLM solve whatever it can and reserve my Claude budget for the rest.

It's waaay better than GLM 4.7 (which was the open model I was using earlier)! Kimi was able to quickly and smoothly finish some very complex tasks that GLM completely choked at.

The old Kimi K2 is better than GLM4.7

Is the Lite plan enough for your projects?

Very much so. I'm using it for small personal stuff on my home PC. Nothing grand. Not having to worry about token usage has been great (previously was paying per API use).

I haven't stress tested it with anything large. Both at work and home, I don't give much free rein to the AI (e.g. I examine and approve all code changes).

Lite plan doesn't have vision, so you cannot copy/paste an image there. But I can always switch models when I need to.

It was the first model I used that was half decent at coding. Everyone remembers their gateway drug.