-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Wiki Selling TSLA Options - Be the House
- Thread starter adiggs
- Start date
ProfessorZoom
Member
Do you personally use FSD everyday? Not trying to invalidate your opinion, just curious
Knightshade
Well-Known Member
Yes.
I've been using FSDb, for years now, on like 95% of driven miles.
it's not remotely close to RT capable, but is a pretty cool L2 system for those 95% of miles.
Also- that 95% would be lower if I drove more often in significant traffic, as a decent # of my disengagements are to deal with behaviors that are only annoying or dangerous with other drivers around (stupid/incorrect lane changing/diving into turn lanes when needing to go straight so it has to jump BACK quickly-- or being overly slow at intersections and stop signs)...
I just got 12.3 today, so I'm hoping that reduces the 5% it works badly on, and handles some of the "with others around' situations better--- but from the vids I've seen it's still a long way from robotaxi.
Last edited:
Using neural nets means progress will be unpredictable, with potential for large leaps in capabilities in certain scenarios as well as regressions in others, but with enough quality training data it should trend to the positive and possibly quickly.Interesting that your experience is different. But replacing 300,000+ lines of code with Neural Net training means progress is going to be very fast now. My understanding is that they are teaching a computer brain to drive just like a human brain would learn.
I'm closing my protective spreads tomorrow morning if we open green. We had a major sell off so far this year on slower auto growth. The growth of Tesla AI and FSD is about to happen. I'm strapping in for 2019 V2!
Neural nets are a black box so you can’t easily tell why it does or does not do things you want. All you can do is feed it more data and hope for the best. But you could easily have a scenario (made up for example purposes) where you retrain it and it now works absolutely perfectly at unprotected lefts but sometimes decides not to stop for red lights and you don’t know why.
I also wish people would stop mentioning the 300000 lines of code thing like that number really matters much. The difference in how it works is significant though of course.
You mentioned Rob’s Chicago video which sounds great, but if you watch Bradford’s it requires several interventions. We are still so very very far from a robotaxi situation. I don’t think people often really appreciate how far it is. But getting people to feel satisfied enough with it as an ADAS to subscribe is likely getting a lot closer, which may have meaningful revenue impact.
I’ve heard this as well, I believe it was mentioned by a Tesla FSD engineer. Unified stack still in the making, until then they are separate (HWY not FSD AI Stack).
TSLA up $4.50 in overnight trading…
Anybody planning to sell into strength tomorrow/this week or wait and see?
I’m ambivalent to sell/short too much since almost all EW technicians I follow are saying the final correction appears complete as an ending diagonal around $160-$158 give or take (max 6.18% is $153.13 and can be SL) and a potential wave (3) should begin as an ABC like the move up for (1) off Jan 2023 (but can also not be a traditional 5-up/3-down). Either way up is what’s being watched. Especially with the bullish divergence on the 2-hr.
Similarly, TA folks are calling out the bottom of the channel @$165-$160 for a rotation back to its top around $210.
Seems not a good time to get too frisky or trade emotionally.
Perhaps we see $180’s but can rotate back down to $160’s and fill in the wedge all next quarter.
Or we just range $160-$175 until earnings.
Anybody planning to sell into strength tomorrow/this week or wait and see?
I’m ambivalent to sell/short too much since almost all EW technicians I follow are saying the final correction appears complete as an ending diagonal around $160-$158 give or take (max 6.18% is $153.13 and can be SL) and a potential wave (3) should begin as an ABC like the move up for (1) off Jan 2023 (but can also not be a traditional 5-up/3-down). Either way up is what’s being watched. Especially with the bullish divergence on the 2-hr.
Similarly, TA folks are calling out the bottom of the channel @$165-$160 for a rotation back to its top around $210.
Seems not a good time to get too frisky or trade emotionally.
Perhaps we see $180’s but can rotate back down to $160’s and fill in the wedge all next quarter.
Or we just range $160-$175 until earnings.
Last edited:
I will probably STO some 3-22 180 CC's
SebastienBonny
Member
I closed 03/29 215 CC last week for about .30 (sold for about 5).
Keeping my hands of selling new CC until a decent leg up and the possibility to sell around the same strike for end of april expiry.
Still holding 220 -P expiring 04/19. Looking to roll it further when we get higher. Rolling 2 months right now is about 1-1.5 premium which could get better when the stock price gets closer again.
The sold put is covered by cash, so no worries getting assigned as well.
Keeping my hands of selling new CC until a decent leg up and the possibility to sell around the same strike for end of april expiry.
Still holding 220 -P expiring 04/19. Looking to roll it further when we get higher. Rolling 2 months right now is about 1-1.5 premium which could get better when the stock price gets closer again.
The sold put is covered by cash, so no worries getting assigned as well.
Just to keep this in mind.I regularly drive with no interventions on 11. So, if anything 12 has not degraded much
One thing we should all keep in mind is - FSD and robotaxi are not what it does. It is about what it doesn't i.e. see what mistakes FSD does in one month or one year (or ten years). For FSD to be robotaxi level, it needs to make no more than one mistake in 10 years of driving.
ps : But here is the important part - at any point the market might decide Tesla has "solved FSD" for no particular reason. So, we may have a Hertz or NACS moment that adds 30% or more to the SP.
With fleet of 400k people using fsd, it needs to have one mistake every 13.4 mins or less to meet you one mistake per car per 10 year.
If you are talking about 1 disengagement every 10 years total, then that makes fsd like 100000x safer after than a human.
If I was really cynical I would proffer the theory that the manipulators are using the weekend price-rises to pop the stock a bit in order to wipe out @BornToFly's puts prior to the FOMC / P&D / ER rug-pull
I think a DCB at best, but WTF do I know?
I have 100x -c170's in play for this week, I'm inclined to roll them until P&D - as always I have a get-out roll to September if needed...
Man I am thinking about closing my calls. Dl004 bull divergence was about fsd apparently. V12 has rave reviews from practically everyone and something like this might get licensed by legacy auto. The first legacy auto that license the thing will give Tesla 1trillion dollars MC statues overnight.
thenewguy1979
"The" Dog
I got +157.5p exp 3/28.
Im holding till after FOMC as this may well be a DCB like Max mentioned.
If TSLA moon this week then lesson learned to avoid long and spread.
Im holding till after FOMC as this may well be a DCB like Max mentioned.
If TSLA moon this week then lesson learned to avoid long and spread.
thenewguy1979
"The" Dog
That should be months out if not years IMO as the tech need to be proven and legacy will take a long wait and see before diving in with all the regulatory fiasco.
That doesn't stop people from positioning themselves for such an event. Tsla usually never trades on fundamentals.
Knightshade
Well-Known Member
There's no such thing as a "disengagement rate" for humans- and suggesting there is misunderstands what disengagements are.
Disengagement rate is not accident rate.
They're the human, performing the OEDR task, recognizing something the system is NOT recognizing that needs to be recognized and taking over full control to correct or avoid the issue. There'd be lots of disengagements, with Tesla not having a complete OEDR, that wouldn't necessarily lead to an accident.
Disengagement rate may well be a useful measure of how much the system is improving from one version to another- but does not, inherently, tell us if the system is safe without a human (short of, as you say, some insanely insanely low disengagement rate far in excess of human accident rate).
If you're talking about an L5 system it's even more complex-- because the existing fleet providing the disengagement rate has a lot of selection bias. A super low disengagement rate in southern CA, where the most concentration of Teslas (and the most training data) is from doesn't necessarily indicate one in rural Nebraska where there's relatively little training data and cows wandering into the road or whatever.
Anyway, we already know from previous data the accident rate of FSDb is lower than just a human-- if that's your goal we've already been there for a while.
The problem is you can't remove the human because FSDb does not have a complete OEDR (or any fallback capabilities that L4 and L5 requires). Don't misunderstand "L2 that keeps getting better" as L4. It's fundamentally not.
Similar threads
