Should I buy this CPO 2014 MS 85

[eye.surgeon]

You're better off with AP1, and in the Bay Area, you're better off without AWD. Elon's promise that AWD was more efficient has never been realized, and it's just another thing to go wrong. Frunk is wayyy bigger in RWD also. Personally I think the early AP1 RWD cars were the sweet spot of the Model S, most reliable and as or more functional than current production.

 

[BrianSF]

You're better off with AP1, and in the Bay Area, you're better off without AWD. Elon's promise that AWD was more efficient has never been realized, and it's just another thing to go wrong. Frunk is wayyy bigger in RWD also. Personally I think the early AP1 RWD cars were the sweet spot of the Model S, most reliable and as or more functional than current production.

thanks for your input, if you don't mind me asking, how's your typical commute like and do you use autopilot function often? i believe my future commute will likely be from the east bay to either the city or peninsula (so lots of freeway driving)

 

[BrianSF]

@BrianSF - I will say a couple things - I have an air suspension car and a coil. I'd go for air every time so good choice on that.

Having said that - depending on your budget and your tax bracket you have a choice here:

AP1 for $56K
AP2.5 brand new for $69.5K after tax - that's the lowest price config with autopilot. For your roughly $15K extra you get:

-glass roof
-significantly faster acceleration
-better seats
-all wheel drive
-roughly 50K extra miles of warranty
-48K fewer miles
-quieter car
-improvements to safety (seatbelt design changes for small overlap crash)
-suspension system improvements
********-a much more capable autonomy platform that's on a very rapid development trajectory and has the entire company laser focused on improving it as fast and as much as possible.
-some improved interior materials
-numerous other design and engineering changes from three years of development.

Thanks for your input and this was the struggle for me too when I ran the numbers on the new 75D. I also don't doubt eventually AP2.5 should be on parity / surpass AP1.

That being said, i'm only eligible for the federal part of the credit (7.5k) and even $56k is really stretching against what I feel comfortable in buying at this moment. I think the plan for me will be to drive this for few years and either upgrade to another CPO S or X in the future (I also hate stomach'ing the immediate depreciation that happens when you drive off the lot)

 

[calisnow]

Oh, puh-lease.

Now see here, Sunshine...

Does *.42 solve the speed limit fail?
How about lanekeeping?
How about the TACC foible at low speed?

I didn’t think so.

And all the cherrypicking in the world won’t change that.

Again - drive a non-vanilla, real world road with both AP2 *.[latest] and AP1.

Newbies have enough sunshine blown up their skirts. Until AP2 is as good or better than AP1, it isn’t.

And it isn’t.

Go drive both cars and see for yourself. It’s a matter of a simple afternoon at Terranea with either of the local clubs.

I own both cars - I fly back and forth between two different states and I keep one in both places. AP2 on .40 and now .42 is a night and day improvement on very crappy real world roads that build .34 and earlier failed. I would grant that prior to .40 there was still a perceptible diff in some situations - especially tar lines, low contrast concrete and high glare.

Here's how to tell if your perceptions and prioties are miscalibrated to the audience of normal people asking for advice: if you go to car club meets to compare performance differences between two systems - you are out of touch and said systems are indistinguishable to "civilians."

BTW you're not gonna trade your car for an AP1 - you're the one blowing smoke lol.

 

[Sunshine State]

I would've agreed until .40 - at this point guys like you and Tao are the equivalent of audiophiles straining to hear minor differences in extremely good gear and losing the forest for the trees. AP1 is a horrible value proposition unless the car has a huge discount. It is a dead end platform in maintenance mode. An excellent platform - but it has no future. AP2 has had lightening fast progress in 12 months - far faster than AP1 did - it drives indistinguishably for "civilians" in the latest builds of .40 and later - and all Tesla's resources are pushing it forward.

It is one thing to have in-crowd nerd fights about almost imperceptible differences between systems (if you're talking about comparison drives at the "terranea" club you know you are out of touch with the real world) among the crowd here - but when strangers come asking us for advice let's try to keep some big picture perspective about where the value lies and where the future is.

Actually I agree, I updated my earlier post to read that Tesla is learning quickly. They realized weaknesses with AP2 and it has been now been replaced with 2.5. We all hope it will eventually get there but in the meanwhile people just have to use caution living with AP2’s quirks. I disagree with AP1 being a horrible value. The best value would be the last of the AP1 cars as they have the most updated cars with the agreed best performing AP. After all everyone wants AP2 to reach where AP1 already is.

 

[calisnow]

... even $56k is really stretching against what I feel comfortable in buying at this moment. ...

Ah - okay then the choice is clear - never stretch your budget beyond what you are comfortable with. Buy the car and enjoy in good health - you will love it!

 

[calisnow]

I disagree with AP1 being a horrible value.

Good point and poor choice of phrase on my part - it's only a horrible value (like anything else) if you pay too much for it. It is an excellent performing system that is still like "magic" to almost everyone in the world except the lucky few.

 

[Sunshine State]

It is a shame Tesla has to waste resources chasing after what they could have been buying from MobilEye if not for personal disagreements. People are arguing about if or when Tesla will catch up to a system designed and built years ago. MobilEye is the undisputed leader in the AP Industry. Tesla should be concentrating its resources on building cars instead of working on matching where they were years ago, it should have been left to the professionals. AP could have been evolved so much further, instead it is either still the same or behind where it was over a year ago.

 

[IAmKai]

@calisnow I'm curious now. You're the first person I've seen defend AP2 this rigorously. I have driven neither, but I've read SOOOO much about how AP1 is generally better, so I was not surprised at all to hear @TaoJones take his stance.

A recent reddit post from an L.A. driver compared AP1 and AP2. He made many assertions, but the one that concerned me during my recent purchasing dilemma (which was very similar to OP's) was:

"AP2 performed poorly on LA Freeways (lanes marked with black-white-black dashes, triple lines at HOV boundaries), seeming to continuously wobble where it decided the lane boundaries were (and thus wobbly steering inputs), and also more likely to fail altogether"

I asked him for details, to which he replied:
"The car's idea of where the lane is will wobble continuously as it drives. It's actually visible on the dashboard display. In terms of steering input, it gives you a kind of uneasy ping-ponging left and right rather than the rock-solid straight line you'll see from AP1. And your odds of a weird steering input, that would send you into another lane, another car, or a barrier, go up to what I would call uncomfortably high. 2-3 times I stopped it from veering left

I didn't mind it while stuck in traffic, since at low speeds, the effect was negligible. It is LA, so that is what I spent most of my time doing. But at normal speeds, it needed a very tight leash. If I wasn't studying the behavior of the system, I'd probably have left it off."

Looking back at the post now, he doesn't mention a firmware version, though.

What's been your experience on L.A. freeways, specifically in the HOV lane?

 

[Sunshine State]

@calisnow I'm curious now. You're the first person I've seen defend AP2 this rigorously. I have driven neither, but I've read SOOOO much about how AP1 is generally better, so I was not surprised at all to hear @TaoJones take his stance.

A recent reddit post from an L.A. driver compared AP1 and AP2. He made many assertions, but the one that concerned me during my recent purchasing dilemma (which was very similar to OP's) was:

"AP2 performed poorly on LA Freeways (lanes marked with black-white-black dashes, triple lines at HOV boundaries), seeming to continuously wobble where it decided the lane boundaries were (and thus wobbly steering inputs), and also more likely to fail altogether"

I asked him for details, to which he replied:
"The car's idea of where the lane is will wobble continuously as it drives. It's actually visible on the dashboard display. In terms of steering input, it gives you a kind of uneasy ping-ponging left and right rather than the rock-solid straight line you'll see from AP1. And your odds of a weird steering input, that would send you into another lane, another car, or a barrier, go up to what I would call uncomfortably high. 2-3 times I stopped it from veering left

I didn't mind it while stuck in traffic, since at low speeds, the effect was negligible. It is LA, so that is what I spent most of my time doing. But at normal speeds, it needed a very tight leash. If I wasn't studying the behavior of the system, I'd probably have left it off."

Looking back at the post now, he doesn't mention a firmware version, though.

What's been your experience on L.A. freeways, specifically in the HOV lane?

My experience is in Florida and I believe we regularly experience what they refer to as the Ping Pong effect going around slow bends at high speeds on highway. The AP2 makes many, many small corrections to negotiate a gentle bend on the highway. My wife says it makes her feel sea-sick after long distance but I honestly think she exaggerates

 

[IAmKai]

My experience is in Florida and I believe we regularly experience what they refer to as the Ping Pong effect going around slow bends at high speeds on highway. The AP2 makes many, many small corrections to negotiate a gentle bend on the highway. My wife says it makes her feel sea-sick after long distance but I honestly think she exaggerates

Ah, I see. Any AP1 experience to compare that to?

 

[calisnow]

It is a shame Tesla has to waste resources chasing after what they could have been buying from MobilEye if not for personal disagreements

Look at the long term view - not the short term. The split was inevitable - the question was merely when. I believe both Tesla and Mobileye did the rational thing to split when they did. There is a plausible argument to make that Mobileye's IP and historical expertise was doomed to irrelevance due to the march of technology. And in fact many Mobileye stock watchers felt the company was doomed - if you go back 12-24 months and read discussions - due to the onslaught of GPU supercomputers powering new learning methods. Also if you look at what Tesla has done - built in 12 short months a driving system that is indistinguishable in the real world to almost all consumers to the one powered by Mobileye's object and path recognition - you can see the naysayers had a point. But 12 months ago nobody knew if Tesla would pull off the miracle they have in fact pulled off. And Intel paid $15 billion dollars for Mobileye. If Mobileye had continued to allow Tesla to use its chips while Tesla developed its own in-house system, then Mobileye's market value could be seriously diminished today vs what it was worth a year ago. Also - Mobileye did slow Tesla down - and that's a good thing from Mobileye's perspective.

From Tesla's perspective the split forced it to accelerate development of its own vision and kick away the training wheels - now it owns all the IP and answers to nobody. Never again can it have the foundation of its autonomous driving system pulled out from under it by a licensing dispute.

If you actually look at the pace of feature and performance releases of AP2 from last October until now it has been breathtakingly fast. Only in the world of cranky laymen does it look like "slow" or non-existent progress.

A company that should be concentrating on building cars

Oh is that what they're forgetting to concentrate on? Of course! :facepalm:

... is working on matching where they were years ago, leave it to the professionals.

AP2 and 2.5 cars powered by Tesla's proprietary vision are still the most advanced, best performing semi-autonomous vehicles available for purchase now by any manufacturer anywhere - with the possible exception of the 2018 MBZ S Class using what Benzo is calling its "4.5" generation autonomy suite. I'd love to see a face-off torture test on bad road conditions between the latest AP2/2.5 build and a 2018 S Class. The Cadillac is not in dealers yet and reports looks like it's out of the running for the current performance crown. Audi's A8 isn't available yet. Drive Pilot II by Volvo isn't competition for the latest AP2 builds - that really leaves MBZ S Class as the sole contender.

If a 2018 S Class shows up on Turo anywhere near me I'd love to get some folks together to do our own torture comparison test.

 

[calisnow]

My experience is in Florida and I believe we regularly experience what they refer to as the Ping Pong effect going around slow bends at high speeds on highway. The AP2 makes many, many small corrections to negotiate a gentle bend on the highway. My wife says it makes her feel sea-sick after long distance but I honestly think she exaggerates

You are correct - it did do the ping pong. What build are you on? It was quite noticeable on early AP2 builds this year. It was diminished significantly with "silky smooth" - almost gone from my perspective on .34 - and now completely gone on .40 and .42 by my "butt meter."

 

[calisnow]

@calisnow I'm curious now. You're the first person I've seen defend AP2 this rigorously. I have driven neither, but I've read SOOOO much about how AP1 is generally better, so I was not surprised at all to hear @TaoJones take his stance.

• AP2 is a really fast moving target - discussions about it are outdated extremely quickly. A report from literally a couple days ago can be false if a new firmware version just dropped.
• Rapid progress can be hard to see unless you track the data. All kinds of improvements on AP2 were happening frequently (every few weeks) even prior to the last 14 days - but this does not smack you in the forehead unless you see it all laid out in one chart or table (I'll attach one to another reply and tag you).
  
• Within the last 14 days or so two new builds were pushed out that dramatically improved performance on the sh*tty roads you mention in your Reddit quote below - the first new neural nets since Andrej Karpathy was hired in late June - and the first customer builds since Tesla began uploading large video files for reinforcement fleet learning in May. The training of these neural nets is an iterative process and it should surprise nobody that it's a rocky process. If you don't have a firmware version in your discussion it's likely a useless discussion. And remember - not all owners have the same firmware at the same time.
  
• This dramatic improvement should be no surprise to any owner paying attention. I am gobsmacked watching the incredibly fast development of AP2 over the last year - starting from zero, to a wobbly toddler - to something that in the last week or so is starting to look like a formidable powerhouse. I keep a spreadsheet that tracks update frequency, new features and reported performance improvements. It's almost scary to watch it iterate this fast. The wider press and public do not see these firmware pushes - you have to look for the thread reports on forums like this one to see what has changed and when. Even then - if you do not actually track the progress data in a table or chart (I do) it's easy to believe progress is slow - until you look at the table and realize that actually the progress is so fast it's frightening (I'll attach it in another reply). The updates started last December every roughly 4 weeks - then 3, then every 2 weeks since Karpathy came on board - now the last two were only a week apart.
• Complaining voices are often the loudest and most distorted and make people blind. But others are really stuck on what it still lacked and gain satisfaction predicting failure. They don't bother to study the rate of progress and instead lazily extrapolate from current performance and Elon's "proclamations" a year ago. Folks are mentally really stuck on the fact that Elon continuously spouts overoptimistic timeframes and others get really mad because they believe he made promises that are unknowable (self driving with the current hardware).
• A few folks have personal axes to grind with Tesla that are hindering their ability to appreciate the speed of the AP2 evolution. These are mainly ethical issues with Elon's early promises and or emotional problems (feelings of hurt and perceived betrayal).
• People falsely interpret Elon's wildly optimistic timeframe promises as indicators that AP2 is moving slowly, or is in trouble, etc. This one is kind of Elon's fault. The trouble is this neural net AI learning is so cutting edge that nobody can predict the timeframes - so laymen believe Elon's stated frames were accurate and take the fact that it is taking longer than Elon stated as a sign of "trouble" in autopilot land.
• People try to read tea leaves and gossip articles on staff changes to interpret the state of progress of a computer science project they know nothing about. I am guilty of this also when instead I should simply look at the data that shows how fast this system is progressing and iterating. Elon's history of publicly throwing out wild timeframes is a legit discussion but it is a different discussion. The fact remains that AP2 is evolving stupid fast - and that has nothing to do with what Elon's mouth throws out on Twitter.
• Weird, totally false press articles are frequently written saying AP2 has not been progressing - when the data is the complete opposite.

A recent reddit post from an L.A. driver compared AP1 and AP2. He made many assertions...

• The two most recent builds - .40 and .42 - have completely changed that equation - but they are so new that large portions of the customer fleet do not even have them yet. I cut out most of your paragraphs for the sake of readability of my reply - but I would say that Redditor's is correct in his assertions about any AP2 build prior to the two most recent ones (.40 and .42) performing poorly on poorly marked and screwy freeway roads. They are literally only a couple weeks old. Tesla has been pushing out new builds ever 3-5 weeks since last December - and the frequency of the builds has been increasing. The last four or so were only 2 weeks apart. Then the last couple only a week apart.

What's been your experience on L.A. freeways, specifically in the HOV lane?

• My experience is that prior to 10 days ago my car did very poorly on the challenging conditions that Redditor described. I had version .34. When my car downloaded .40 however this situation completely changed and my car suddenly started driving like a freight train on rails even in these garbage conditions. Los Angeles is a wonderful torture test and I think that's one of the reasons Tesla's AP1 (and now AP2) systems seem to be the world champions of great autonomous performance on bad road conditions - because instead of the Europeans their neural networks are trained on some of the sh*ttiest roads anywhere! But don't take my word for it - read the incoming reports of other drivers. I'll link:
  • Just got new update: 2017.42 a88c8d5 - user @ecarfan speculates .42 may be using the crowd sourced HD maps Tesla has touted because his car is now slowing before road bends can even be seen
    
  • Just got new update: 2017.42 a88c8d5 @conman's report on winding roads of .42 - much improved performance.
    
  • Wow - AP2 HUGE improvement with .40 firmware! - My own giddy thread from a few days ago reporting my dramatic performance improvements on getting .40: "I'm delighted! First chance to drive .40 through tough freeway conditions that tripped up .34 and earlier. It's smooth sailing for me through: construction cones on I39 south to Madison. High glare combined with mottled shadows over fresh tar lines! This was so crazy it was undriveable before. Now almost perfect!" "Wowwwww baabbbbyyyy!!!!! Sun dipping low - super high glare now - concrete not asphalt - light gray with white center markers. Torture test that AP1 failed for months when it came out. .40 is SOLID AS A ROCK! Thank you Mr. Karpathy!" "High speed left hand sweeper - 70 mph - interchange highway to highway - right hand lane marker nuthin' but a muf*ckin' tar line - high glare - concrete - annmddd weeee diidddd itt!!!!!"

• The other thing you see on all forums is the human tendency to vastly exaggerate problems with consumer products because people get psychological satisfaction from b*tching. You can find folks complaining about $50,000 Sony home theater projectors, $1,500 Austrian hand built Kastle skis - and now AP2. Usually the loudest complainers don't understand the technology, don't understand the iterative learning process of neural networks, and have personal emotional axes to grind with the corporation or founder (Musk) himself. There are a lot of people who seem to have emotional wounds like you'd see in a romantic relationship. Some folks around here could have their Tesla sprout wings and they'd still focus on the fact that they feel they were lied to. It's sad to watch them.

***I must say as a disclaimer however that my AP2 car is "stationed" in the northern Midwest with me (Wisconsin) currently and my AP1 car is in So Cal. I obviously can't drive it on exactly the same "garbage" as that Redditor but we have plenty of "garbage" up here to torture test on as well.

 

[calisnow]

@IAmKai here is a screenshot of my simple autopilot dev/update tracker (click the image to expand so you can actually read it). You see the left column highlighted - the number in each cell is how many weeks passed since the last update. Notice that number keeps shrinking. . .

 

[calisnow]

Here is a graph I made and posted in another thread a few weeks back which shows that the frequency of software updates has been steadily increasing (the time between updates is shrinking).

 

[calisnow]

You're better off with AP1

If you haven't driven .42 AP2 please stop telling people this because you're spreading misinformation - your source material is outdated.

 

[JPUConn]

If you haven't driven .42 AP2 please stop telling people this because you're spreading misinformation - your source material is outdated.

Glad you included this version note. I drove ~1k mostly AP miles 2 weeks back on a loaner AP2.5 Model X with .38 and it was beyond awful. Getting back into my AP1 after was a leap forward. I’ll have to ask to borrow on with .42 on it.

AP1 - fantastic for 99% of real world commutes. Sure it could handle merging lanes and single lane side roads better but I’m ok with it not taking off ramps and accepted it won’t do traffic lights or stop signs. I’ve happily accepted that this may be the best this hardware version gets and I’m ok with that.

AP2/2.5 (version .42) - improving on feature parity and reliability like AP1 and eventually will have much much more to offer.

I don’t keep cars very long so AP1 made sense to me from a value position. Next one I’ll get AP2.

 

[Sunshine State]

I’m obviously a big fan of Tesla shown by the fact that I have a great amount of money in Tesla cars but I don’t let that cloud reality. Those who argue that AP1 was maxed out should realize MobilEye, with their far greater technical resources would have continued to advance it. It is common belief that technology doubles every two years. Tesla might be catching up to 4 year old AP1 technology now but that should not be considered a great feat. That by normal standards should be a disappointment as they already had the existing system to learn from. Tesla claimed development was going to be much quicker, if they knew what they were doing was that claim know to be false? There is no reasonable argument that AP1 would not have been so much further ahead of where AP2 is by now had development been continued. There was no benefit to Tesla to improve AP1 other than fulfilling their promises and that is not always a priority for Tesla.

 

[Zorg]

Personally, I am shooting for either S85 with AP1 around $50K (I know, may be wishful thinking) or a 3. The S is bigger even if it has less range, but it would allow me to drive from the east bay to Tahoe in the summer if need be, and more than sufficient to commute.