[benjamincburns]

I'm curious to know how much work they'd done on the idea ahead of time. If this was conceived during YC, they're definitely a very impressive team.

A few months ago I looked into building this exact product. If you're a DIY kinda person and you don't mind carrying around some extra bulk, it's somewhat easy to cobble something together. If you want a light/small brick w/ self-contained charging and output regulation that just works, it's much more difficult and potentially dangerous -- especially if you want it to charge fast enough that a 1h airport layover is meaningful. Add in more advanced battery management/conditioning schemes, and I'm just all kinds of impressed.

Though, this claim does irk me a little:

> With the same battery, we could charge to a full 100 percent and it won’t degrade

> Never experience loss in capacity again. A Li-ion battery powered by BatteryOS doesn’t degrade in capacity over time, and as a result has a 4x higher cycle life than when controlled by conventional methods.

I think I need a better definition of "capacity degradation" here - if you acknowledge that there's still a charge cycle limit, I'm curious to know what happens once it's reached. Either way, I'm very sceptical of anything which claims to completely eliminate the effects of entropy.

[esmi]

There are 100s of ways to game this cycle spec too. For example, for all we know there is a 100WHr pack inside the box and they only make 50WHr of it available to the user. This way you can make a 50WHr battery that effectively never "degrades". The BatteryOS just manages the "reserve". Then the only thing they are really measuring is the speed at which their algorithm looses calibration. The software for many BMUs (battery management units) actually does this today.

[benjamincburns]

That's not a terrible idea, actually.

When you purchase any NAND flash device like a USB stick or an SSD, essentially the same thing is happening. The raw storage is much higher, but due to a high raw bit error rate (see NAND displacement errors), the need for wear leveling, and the need for reserve blocks to swap into service as the device ages, the claimed capacities are reduced to accommodate MTBF and observed bit error specs.

In cases such as this I think it'd be disingenuous not to drop your claimed specs to reflect the actual usefulness of the device.

[nardi]

Except they explicitly state that they don't do this. They refer to the 50% charge scheme being used in the Chevy Volt and claim they don't do that.