FSD SoCal - ₸⚡🚘🔋🚥

Tesla FSD Tester in Southern California. Software Engineer. Dad. Tesla fan and investor. Fair but firm beta tester.

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

@teslascope You are correct once again @teslascope ! Came in last night like they were monitoring this thread ! 😃

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

50%! Sadness. Nothing here. @teslascope is probably right and on top of things. But if so then why is TeslaFi stats show 5.9%!

Is there that few AI4 vehicles?

[Teslascope @teslascope]

Can confirm!

With this wave, we’ve now surpassed the 50% mark for rollout of Full Self-Driving V14.2 to eligible AI4 vehicles.

This has been one of the fastest rollouts of a singular FSD build, and more waves are expected to follow soon!

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

Woot! Can @Tesla_AI take a pic of the whole puzzle for us ?

[Elon Musk @elonmusk]

@DevinOlsenn 14.3 is where the last big piece of the puzzle finally lands

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

Wowsa. FSD 14.1.4 seems comically bad in the rain. It's coming down pretty decently in SoCal and I got 20+ take control immediately red hands of death in a short 5 mile trip. Pretty unusable. Cameras were partially obscured with rain (duh) but v13 handled this way better

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

Yours is a fair take especially with the “FSD is solved” language repeated. I get where that's coming from with 14.x future path, but it's wishful thinking unless proven otherwise.

Report the issues! Every version is better (mostly) but it’s a journey to true FSD.

It does make me super happy with Robotaxi existing, meaning any issues now have to be front and center of the AI team

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

@OwenSparks @amy787 @aelluswamy Also the front facing camera is taller than most human eyes. I am 6ft for example and it it’s higher than me

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

@JonBbC_TechGeek Somehow I got v12 in a small group before the OGs. I recorded videos that were balanced and included all issues. It’s not that hard. Tesla is not your mom, nor is Elon your dad. It’s a company trying to get better and needs your disengagement data to do so

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

@TeslaTony310 @TeslaTruckClub In which way are you referring to?

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

On a fun note. FSD 13.2.9 was smooth and good enough to pass the “wife test”. I think 14.1.3 and onwards continues to meet and improve that bar (mostly) and the smoothness wrinkles with 14.x will get ironed out.

Still disengagements though. Let see what 14.2 brings!

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

FSD 14.1.3 Just had 2 disengagements in a 5 mile drive. Not critical.

Going to rant not against Tesla or FSD but the common community here. You are not doing anyone favors (or Tesla) by “letting it figure it out”. Disengagements for bad behavior is what improves FSD.

Don’t come with bullshit like “video or it didn’t happen”. Tesla has the video. You on the other hand ignored previous videos of similar behavior and always praise the next version as perfect. /rant

Overall this early version is better and will recover from failures most times… BUT GET TESLA the data dammit!! @Tesla_AI

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

@TeslaTruckClub 7 years for me. I am saying TSLAQ and others (Dan O) has influenced attitudes and content here even if most of it was years ago. Plus people literally get paid in here and want to be on Elon and the communities good side.

That is a recipe for where we are today.

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

Exactly.

To be fair we all should strive to be balanced. But otherwise report the bugs then things get better.

I get that some are probably jaded due to TSLAQ and negative nellies taking things out of context. But unbalanced leads to a different loss of trust.

14.1.3 is great overall off to a strong start as an early release. Break stabbing or odd behavior doesn’t happen much (to me). But parking logic / nav (a marquee feature) and U-Turns need help. My list of my bugs is shrinking overall though!! @Tesla_AI

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

@matthewtalmage @Teslaconomics A bit of semantics but there is a Google paper that proves LLMs learn in real time context window replacing underlying model weights temporarily. Similar thing could be happening here with vision input / tokens.

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

(3/3) FSD 14.1.3 first impressions

Unchanged: U-Turns still have issues. #1 Still not executing required U-turns. FSD had no lead car and blew past 3 chances to make a U-Turn. RoboTaxi would have likely gotten stuck or remote call trying to get out. Tesla still has not fixed despite many videos and reports.

#2 - Slightly improved U-Turn onto 1 lane (sharp turn) but not like a human would. Car behind me doing the same actually cut in front during the maneuver due to the suboptimal path FSD took. Humans execute wider turn (with no car in right lane) to make it 1 smooth shot.

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

(1/3) My v14 first impressions. Nice improvements but not yet RoboTaxi ready for my area/routes.

Good: What other reviewers have noted. Assertive. Quick decisions. Hurry actually hurrying and Mad Max almost like in video game.

Bad: Some harsh braking at protected yellow right turn arrow when in the middle of it.

[FSD SoCal - ₸⚡🚘🔋🚥 @FSDSoCal]

(2/3) V14.1.3 impressions

Meh: parking was meh for me and had to disengage with 2/2 attempts. For first, Pin was correct but route wanted to go through no drop off zone.

For second parking it did not want to give option for street parking at location (which is closer) than parking lot. I put pin exactly at location but it didn’t change behavior. FSD wanted to blow past preferred parking and walk would have been longer