State of Charge with Director of Cost Engineering Corey Steuben

[borski]

Yeah, one great example of this is OpenAI. Previously, they found that by simply throwing more resources at the problem allowed them to improve ChatGPT, then came GPT-5, which didn't give back improvements despite getting so much more. I could go into deep lengths as to why, but long story short, it's the inherent problem with the architecture that they've hit such a point of diminishing returns.

Bringing it back towards automotive, you can only use "throw more things at it" for so long before you've hit a wall

I completely agree. While more resources can resolve some problems, at some point you do hit diminishing returns, as more resources cannot and does not resolve all problems.

And logic bugs are logic bugs regardless of resources.

 

[Need4Sleep]

I completely agree. While more resources can resolve some problems, at some point you do hit diminishing returns, as more resources cannot and does not resolve all problems.

And logic bugs are logic bugs regardless of resources.

Especially these days with so many cars coming with ADAS, driver assistance, lane centering, etc., it's becoming ever more important to have an efficient and strong software architecture so you're not held back by latency. We can also an example from Lucid. I recall during an interview with the engineers, they can change damper behaviour based on whether they've detected user input or not. Another example is the fact that Sapphire's traction control is operating at 1000 Hz (if I recall correctly), so for the tri-motors in particular they can quickly change the torque at the rear based on whether it's meant to stabilize the car mid-turn or to go into a controllable amount of oversteer to let the rear end loose.

Savagegeese made a great video covering the software side

 

[chork]

The video felt a little like smoke and mirrors when talking about the Cosmos, mostly when it came to distance the car can go. Steuben did say (at 18:54 in the video) final battery pack size and range of the vehicle achieved was not going to be revealed at this time. At 7:32 in the video Steuben mentions trying to strive for efficiencies up to 4.5 m/kWh. Does that mean the 100% charge is 4.5 x [battery size]kWh or you could hit the 4.5 m/kWh in a perfect driving condition (flat stuff/55 mph, do tell) or...? At 18:06 in the video Steuben states "we're able to go further with less" and mentions going 300 miles in the car you would need a 69 kWh battery. Hmm? Well I can tell you 4.5 m/kWh x 69 kWh = 310.5 miles. Let's hope that the final reveal shows either increased m/kWh or battery pack size or both. Of course, no mention tire size, motor configuration (AWD vs. RWD) to get to this 300 mile figure via 69 kWh.

Also, it was not revealed if this new efficiency that the Cosmos will provide also improves upon battery degradation over time. My leased 2024 Lucid Air Pure over 20 months has experienced 18 mile loss at 100% charge, i.e. 408 miles new to 390 miles now. Also, no mention of an SSB.

Guess we will see what comes out at the end of 2026.

 

[borski]

My leased 2024 Lucid Air Pure over 20 months has experienced 18 mile loss at 100% charge, i.e. 408 miles new to 390 miles now.

This is not the correct measurement of degradation. For example, if your battery is unbalanced at all, letting it sit untouched for 72 hours at 100% SOC would potentially increase your “range.”

The API has a battery health metric, though we don’t know how it’s calculated either.

 

[erigion]

Also, it seems like li-on batteries generally degrade faster during the first year or so then stabilize for the rest of their lifespan.