@mikepat711 I got Cursor Ultra this weekend after using up Grok Bot compute on lower plans. I defer a lot of optimization to LLMs themselves, asking if they think they’re set up well or need adjustments. Then I correct them when assumptions are wrong or they’re clearly missing something.
24 hours in with FSD v14 lite, here are some more of my thoughts/review:
It was worth the wait for sure. This is such a revolutionary change for HW3 and it feels like I have a new car now.
First off, the highway performance is outstanding. Truly something special they’ve created here that really feels like my AI4 cars. The lane changes are truly what makes it feel like v14. Intelligent choices as well as the super smooth lane changes. Immediately noticeable AND really welcome change.
The lane change logic on this build makes very precise and calculated decisions too. So much better at deciding which lane it will change lanes into without making excessive lane changes. Speed control has been good too especially in hurry. Really good at matching the speed and vibe of other cars as well. It was so confident on the highways.
On city streets, the increased reaction time is immediately noticeable. I was driving a few hours ago and it could tell that the car to my right was going to need to change lanes when their lane was ending and it quickly and immediately changed lanes before I could even process it. It is so cool to me how Tesla was able to make HW3 this capable and smart even though it has about 1/8th the memory bandwidth as AI4. Absolutely wild how good it is now.
Lane choices are also way more intelligent, especially nearing exits. It is definitely more cautious as Ashok mentioned for now and on some residential streets a bit slow and sometimes hesitant, but not bad. FSD v14 lite is quick to react if someone opens their car door. The speed bump and dip detection is great, even if they’re hard to see occluded by shade or a hill.
One thing I noticed with this build is that it always turns into the first available lane. This is great and it means the turns are always tighter. This is something I would like to always see in V14.3 and have mentioned but it sometimes goes into the second lane. Sometimes with v14 lite on a tight turn the rear wheel/tire gets close to the curb but never too close yet. I’d love to see a bit of refinement here, as well as merging with a bit more throttle input.
I am definitely enjoying the new UI changes. Really feels like V14 in the AI4 cars and I love being able to press a button from park and start FSD. The difference with my Intel hardware 3 car is that I have to quickly tap the brake pedal to confirm FSD, not a big deal but hopefully this changes.
Handling VRUs well, definitely cautious and playing it safe for now, no hard braking, it always gives right of way. Tuned pretty well now for the first build and will only get better as it progresses.
Parking for the most part has been pretty good. It isn’t too slow usually and it has been parking pretty centered most of the time. FSD v14 lite spot selection choices have actually been quite good so far and it is very quick to decide. Would like to see a bit better initial approach angle to make the maneuver always a bit more linear and quick. If it can set up its approach angle, all of the parking jobs would be always more centered and a bit quicker. What is great is the parking spots so far have always been convenient spots either near a door or the first available if it is a busy lot. Also, the nag has been relaxed over v12.6.4 and is way more enjoyable to drive.
It did great on canyon roads as well. Did almost an hour of very tight, twisty, and dark canyons in Malibu at 4 o’clock this morning. Took good lines, great throttle and brake application points, and was pretty quick unless it got tight then it was a bit slower than AI4 Mad Max.
I’ve now been driving FSD v14 lite since last night almost nonstop between filming, driving, writing these reviews and editing videos. Even though I still haven’t slept- it’s been a fantastic day and I’m so grateful to be doing this.
Overall, v14 lite brought new life to Hardware 3 and so impressed and thankful @Tesla_AI was able to accomplish this. Can’t wait to see how it gets even better in the future!
Most people think the AI buildout is just a couple of stocks – Memory, CPU, GPU, NeoClouds, Photonics..
$SNDK, $MU, $ARM $INTC, $AMD, $NBIS, $LITE, $AAOI, $TSEM, $SIVE
The reality is there’s so much more to it.. I built a complete map of the AI buildout – at least the one I’ve been trading around for the last year.. I put that map on my Substack… 12 layers deep.
Since I’ve seen it circulating on here, I’ll just post it myself.
The hard part isn’t knowing the full stack – it’s making sense of it.
Why is photonics down today but memory isn’t? Why is CPO so relevant in the AI buildout?
Bullish AF on AI infrastructure.
Go check out the full article for free on my Substack, and if you’re going to share it – please do, but link back to my article.
Here is the link: https://t.co/AqKTdxnlmJ
Let’s go find the next 1000% stock ⚡️⚡️
EV ALMADAN ÖNCE KESİNLİKLE BAKMANIZ GEREKEN SİTE
Hangi saatlerde güneşin binaya nereden vurduğunu gösteren mükemmel bi site. Ev alırken, kiralarken mutlaka bakılması gereken şeylerden.
@jcubhilton I brought three sets of dumbbells (10-20lb) to work so I can do mini workouts during the day. I started using my 30 lb dumbbells at home. I have access to two gyms too. I’m not giving up on our muscle up dream, we got this 💪