When AI meets the real world. FSD tester and progress tracker, Builder of @CamCutApp dashcam manager and telemetry visualizer. Tesla fan

Los Angeles, CA
Watch FSD react to an obstacle revealed at the last moment. Larger box: 3/3 avoided
Smaller box: 1/3 avoided We also ran a control test with the lead car swerving without an obstacle. FSD didn’t react, suggesting it was responding to the obstacle rather than simply following the lead car’s maneuver. The smaller box was still large enough to potentially cause damage. @OGMarsLife @Muzeishen @Tesla_AI
5
11
70
3,614
Being on X is literally living on the cutting edge of technology. Absolute insanity.
3
1
20
1,454
Should FSD have swerved to avoid the box? @Tesla_AI
9
1
32
4,694
Btw there is only one correct answer
3
1
528
FSD allows cyclist to safely cross, while human driver has to lock up their brakes to slow down. #tesla #fsd
5
5
26
2,169
I guess this guy didn’t have the new Automatic Collision Evasion enabled 😐
BREAKING a @tesla has crashed into the fremont factory showroom due to driver error no one was injured
4
1
20
3,829
FSD anticipates getting cut off and reacts early like an experienced driver.
8
4
170
20,736
Here is my solution to this edge case If(sees emergency lights AND hears siren) then (pull over) else (keep going); @tesla_ai Did I get the job?
Tesla FSD v14.3.10: Emergency Vehicle Detection Still Needs Work While running a few errands tonight, I hit back-to-back false positives where FSD incorrectly detected the setting sun as an emergency vehicle. Both times the car signaled and started pulling over to the right even though there was no emergency vehicle in sight. I had hoped this would be fixed in FSD v14.3.10, but it looks like there’s still more work to do. Anybody else experience similar behavior w/ v14.3.10?
6
14
3,828
SW + HW The next chapter…
1
1
21
1,440
Driving in LA traffic every day, you really get to appreciate FSD’s ability to look ahead and behind and all around at the same time. Two right-turn lanes here. The car next to me is slightly behind me, outside my peripheral vision, and starts drifting into our lane mid-turn. FSD sees it, calmly moves out of the way, then accelerates and slides right back into the lane once it’s safe. 
No overreaction. Just a really smooth and safe response. 360 view coming to @CamCutApp in the next few days to better visualize these situations
10
12
209
16,387
Here’s why right turns still aren’t solved. From time to time, FSD has a brain freeze when creeping out from behind a blind corner. Instead of slowly peeking out until it can see, it casually just goes for it. There’s an approaching car completely hidden behind the wall here. FSD has no idea what’s coming. @Tesla_AI
77
17
373
57,421
Edge Case retweeted
FSD v14.3.10 avoids a giant mattress blocking the freeway lane.
42
76
836
143,954
Gonna drop this sick 360° dashcam view for all the chilren to play with in the next @CamCutApp update. @MartyEarthy gave me the idea about 6 months ago when I first started building CamCut, but I really disliked that Tesla doesn’t provide the wide-angle front camera feed, so the transitions between the front and pillar cameras can’t be seamless. You’ll have to live with the pillar gaps because I don’t want to fake it, but honestly, it’s not too bad. Kinda makes you feel like you’re sitting in the cockpit of a spaceship Let me know if you think of any feature enhancement or how you would use it.
12
3
29
3,918
This is really interesting situation FSD v14.2 lite has to stop in the middle of the intersection to let emergency vehicles through and then gets stuck there on a red light. I think FSD had to start creeping forward as soon as the ambulance passed to show intent otherwise the cars will start moving. then you’re stuck there for the whole red cycle. Would’ve been worse and more dangerous situation on a much bigger street with fast moving traffic
Did FSD do the right thing here, what do you think? Filmed w/ FSD v14.2 Lite HW3 on Tues. Sept. 15th. @tnertz @edgecase411
2
20
4,068
14.2 lite edge case. Can’t see the construction pole flag guy from far away
TechGeek Tesla 🔋⚡️
3
38
2,268
Tesla owners need to get comfortable with an uncomfortable possibility: Sleeping in the back seat of your personal Tesla may still be many years away. I’ve been tracking FSD progress for almost 4 years. Every major version keeps getting better, but the gains are getting smaller. v11 +1.37 v12 +1.39 v13 +1.09 v14 +0.78 Since v12, each major has added roughly 75% of the improvement of the one before it. If that trend continues: v15 → ~8.8 v16 → ~9.2 We’re already past the average 322-day (10month) gap between major versions, so I’m expecting v15 soon. 8.8 is the number I’ll be watching. If v15 lands significantly below that, it means the improvement curve is flattening faster than the current trend predicts. But here’s the problem: the last 10% isn’t just another 10%. Comfort is already at 9.3 on my tracker. What’s left is a mix of stubborn everyday problems that somehow still aren’t solved, like navigation, potholes, parking, signs, and the long tail of less frequent safety-critical edge cases like thin barriers, overhanging objects and unusual road situations. Navigation is especially frustrating. After almost 4 years of tracking FSD, it doesn’t feel like it’s getting much better. This is the march of nines. The closer we get, the harder each additional nine becomes. And there’s another part of this I think people are overlooking. If Tesla can continue growing FSD subscriptions simply by making Supervised FSD better and progressively less actively monitored, where is the urgency to push personal vehicles into true unsupervised operation? That next step brings a completely different level of regulatory scrutiny and liability. Meanwhile, Tesla has another path. It can roll out unsupervised Robotaxi city by city, control the fleet and operating area, and expand as the system proves itself. That’s a much more controlled way to take on the risk than allowing millions of privately owned Teslas to drive around without anyone in the driver’s seat. So I absolutely expect FSD to keep getting better. I expect the nags to keep disappearing. And I expect Supervised FSD to get incredibly boring in the best possible way. But “better FSD” and “I can sleep in the back seat” are not the same finish line. I don’t know about you but I love this journey. It’s fascinating to experience and I look forward what the next couple of major version releases brings.
95
35
387
35,163
If you want to play around with the interactive tracker and explore all the categories and ratings edgecaselabs.app/fsd/
2
1
22
1,734
I’m going to go out on a limb and say everyone using FSD is experiencing this frequently and is annoyed as hell. Everyone is praying V15 finally solves the late-staging navigation issues so we don’t have these YOLO cut-throughs to make an exit. Then we can finally say FSD is boring as hell, in the best possible way.
Here is the rest of the FSD exit party…😑
144
21
965
229,760
Caution: Do not try this FSD test at home, kids! Testing FSD’s response to an elevated obstacle: a 2x4 sticking out directly into its path. At 30 mph, FSD does not react ❌ At 20 mph, FSD swerves to avoid it ✅ Shoutout to @Muzeishen for trusting his life in my hands and @OGMarsLife for being a witness that no harm was done This test was a bit spontaneous and will be improved in the future. I know that FSD still has trouble with suspended obstacles. But to be fair the wood was not clearly visible to me until about a couple of seconds before I intervened
33
14
276
59,880
Balls of steel man pov
2
15
2,536