Menu
Deep ResearchPROAsk Elon
Elon Musk · Tweet Archive

The tweet archive.

15 years of Elon, fully searchable. The production archive uses Supabase as the source of truth, with 94,952 indexed tweets available in development as a full-archive fallback and a curated annotation layer for context, theory, and how major claims aged.

Showing 1,601-1,650 from the Supabase archive
Jan 7, 2021

@teslaownersSV Well, back to work …

108.7K likes12.9K RT2.3K replies
Jan 2, 2021

@newscientist Um, we have giant fusion reactor in the sky that works with no maintenance

46.1K likes2.0K RT1.6K replies
Jan 2, 2021

@comma_ai Tesla Full Self-Driving will work at a safety level well above that of the average driver this year, of that I am confident. Can’t speak for regulators though.

18.3K likes801 RT578 replies
Dec 30, 2020

@RationalEtienne @biogirl09 @ErcXspace Legs would certainly work, but best part is no part, best step is no step

3.0K likes156 RT129 replies
Dec 28, 2020

@kimpaquette Just meeting with Larry Ellison to seek some advice. Back working on Tesla end of quarter tomorrow.

7.6K likes251 RT521 replies
Dec 21, 2020Viral spike

Thanks to everyone who worked so hard to make Tesla successful. My heart goes out to you.

298.8K likes9.2K RT6.9K replies
Nov 28, 2020

@ID_AA_Carmack Int8 for almost all heavy lifting in inference works great

800 likes16 RT34 replies
Nov 17, 2020

@Erdayastronaut @CharlesNOtrumps @rweb11742 Our first Merlin engine design was really bad tbh. Nothing worked well & some parts didn’t work at all. 18 years later & thousands of upgrades make it a great GG cycle engine.

2.3K likes72 RT61 replies
Nov 13, 2020

@austinbarnard45 Burst disk worked, so vehicle appears to be ok. We’ll have to swap out at least one of the engines.

3.6K likes175 RT141 replies
Nov 11, 2020

@Kristennetten It is an honor that they would choose to work with me. I strive to ensure that the companies enable creative & driven people to reach their potential.

3.3K likes176 RT162 replies
Nov 11, 2020

Such a privilege to work with people I like & respect so much. I feel blessed.

174.8K likes5.3K RT3.0K replies
Nov 1, 2020

@Erdayastronaut @KlotzAdam @NASASpaceflight We’ve tested a sub-scale version in a wind tunnel with active aero closing the loop for stability, so it will probably work at scale, but reality tends to bite you on the ass

5.1K likes197 RT198 replies
Oct 31, 2020

@TesLatino Thanks to a great Autopilot/AI team working super hard! Doing “neural nets” without testing them against reality has led some groups to overestimate what they’re doing. Reality is the hardest sim.

5.9K likes186 RT152 replies
Oct 28, 2020

@jgrano305 @flcnhvy To the best of my knowledge, literally zero equipment was usable without major repairs or upgrades. Most of the factory had no equipment at all! Even light bulbs were missing in some areas. GM & Toyota took everything that worked & moved it to their other factories.

1.6K likes61 RT98 replies
Oct 27, 2020

@HyperChangeTV Tesla team has done great work!

2.3K likes77 RT108 replies
Oct 23, 2020

@bart_sanders Working on it. Also needs to be able fart at other cars on command.

3.0K likes128 RT237 replies
Oct 11, 2020

@Teslarati Great work by SpaceX valve engineering & production!

4.4K likes173 RT109 replies
Oct 9, 2020

@ajtourville Yeah, it’s not a contest. Advanced casting crushes best-case 3D printing, but casting doesn’t work well for complex enclosed volumes & high strength or high temp alloys.

1.3K likes73 RT69 replies
Oct 7, 2020

@flcnhvy @teslaownersSV @raytech247 @WholeMarsBlog That would be like licensing Battlestar Galactica. Actually building continuously evolving gargantuan factories is the hard part. Lots of people who worked at Tesla Gigafactories now work at competitors, but none of those companies have made one.

2.0K likes118 RT113 replies
Oct 2, 2020

@AdlanBogatyryov @Erdayastronaut @NASASpaceflight @BocaChicaGal @TheFavoritist Great letter! It will be increasingly difficult to see Starlink satellites, as we’re actively working with the astronomer community to ensure that even the most sensitive telescopes are fine & scientific progress is not impeded.

2.1K likes114 RT54 replies
Oct 2, 2020

@Erdayastronaut @NASASpaceflight @BocaChicaGal @TheFavoritist The ship rings are thicker than they need to be (for now), so same thickness works for booster & ship for hoop stress. Booster lower tank will have longitudinal stiffeners to prevent buckling.

2.6K likes113 RT84 replies
Sep 26, 2020

Thanks Tesla Team for great work on deliveries! For new owners, we super appreciate accommodating us on delivery timing! Definitely one of our toughest quarters in global logistics.

47.7K likes2.0K RT1.4K replies
Sep 26, 2020

@NASASpaceflight @austinbarnard45 @PPathole 8 bar differential in ullage, 9 bar at base due to propellant head. It’s enough. Improvements in work.

2.0K likes130 RT61 replies
Sep 26, 2020

@PPathole We just need enough height to test body flaps & drawing propellant from headers vs main tanks. Will do several flights to confirm working well, then add heat shield & go high Mach.

2.9K likes163 RT78 replies
Sep 21, 2020

@soumiksf @ID_AA_Carmack Yeah, we will open Dojo for training as a web service once we work out the bugs

1.5K likes112 RT48 replies
Sep 21, 2020

@samiralyateem @ID_AA_Carmack There are a lot of pointless zeroes in FP32 neural nets. You can chop off 16 of the 32 bits without losing meaningful precision. This works well with neural nets, but not with regular computing, which expects extreme precision.

515 likes21 RT27 replies
Sep 21, 2020

@ID_AA_Carmack For Dojo, we’re betting on FP16 with floating floating point (essentially, truncating mostly zeroes), so it feels like FP32 & existing software just works. We considered FP8 for a while, but everyone hated it. I even asked Twitter! For inference, int8 works great.

2.4K likes114 RT79 replies
Sep 19, 2020

A prior track record of exceptional achievement in engineering *is* required, but *no* prior experience working on the brain or human physiology is required. We will teach you what’s known about the brain, which is not much tbh.

20.2K likes901 RT1.0K replies
Sep 19, 2020

@teslaownersSV @Model3Owners @TrungTPhan Felt I had done enough on the Internet (Zip2 & PayPal) that my work would be useful to others. Needed to get going on accelerating sustainable energy & making life multiplanetary.

3.3K likes272 RT166 replies
Sep 19, 2020

@Model3Owners @TrungTPhan It took an utterly insane amount of work to move the SpaceX & Tesla success probabilities above ~zero

3.8K likes321 RT182 replies
Sep 15, 2020

@ValaAfshar Tunnels are the solution to traffic. You can have as many levels as you want. Usable tunnel depth far exceeds tallest buildings, so would work even for New York or Beijing.

11.0K likes837 RT861 replies
Aug 30, 2020

@SamTalksTesla @Teslarati If some company has solved a useful thing that we can just buy, that would be great! It’s so much work to create our own tech. We develop new tech, because we have to, not because we want to.

2.0K likes139 RT136 replies
Aug 30, 2020

@tlowdon @EthicalSkeptic Antibody half-life for cov2 is very short. In a study that SpaceX did working with Harvard epidemiology, it appears to be as low as 2 or 3 weeks. Nonetheless, reinfection rates are very low & symptoms are minor, perhaps due to memory T cells https://t.co/z9WlvIjoTw https://en.m.wikipedia.org/wiki/Memory_T_cell

932 likes111 RT47 replies
Aug 30, 2020

@kulpability @tlowdon @EthicalSkeptic One person at Tesla (out of 60,000) was hospitalized in serious condition (he caught cov2 at home, not work), no deaths. I called his wife & Hayward hospital to make sure he was getting the right treatment, but which he was.

2.2K likes221 RT51 replies
Aug 29, 2020

@PPathole @IEEESpectrum @neuralink A nightcap would probably work best

1.2K likes36 RT142 replies
Aug 27, 2020

@WholeMarsBlog We were too obtuse for our own good, although I think it works esthetically. Those are folded-over current collectors at the top/bottom of the cell, which are important, but I don’t want to jump the gun on Sept 22 …

3.2K likes167 RT125 replies
Aug 26, 2020

Live webcast of working @Neuralink device Friday 3pm Pacific

39.9K likes7.4K RT1.9K replies
Aug 25, 2020

@jchybow @jwangARK @wintonARK It’s almost working reliably

1.3K likes55 RT91 replies
Aug 25, 2020

@jwangARK @wintonARK Very different computers are needed for training vs inference. FSD is mostly Int8, but Dojo is de facto FP32 (feels like FP32, but actually FP16 with boring bits truncated) so we don’t need to rewrite all the training software. It should work almost immediately.

986 likes74 RT39 replies
Aug 19, 2020

@cnunezimages @SpaceIntellige3 I’ve been working in Boca since Sunday. August heat & humidity in South Texas is intense! If you think of it like a tropical resort, that makes it better though.

2.8K likes108 RT148 replies
Aug 19, 2020

@Teslarati We really need a bigger one, so adults can squeeze in too. Guaranteed to make a party more fun. Works well with flamethrowers!

5.4K likes196 RT283 replies
Aug 19, 2020

@krystof_mitka @BiIndia Winning “worst person I’ve ever worked with” is not easy. Some real champions are vying for that title!

682 likes23 RT38 replies
Aug 18, 2020

@C_Bass3d @Erdayastronaut @NASASpaceflight @FelixSchlang Very impressive render. Note, legs will be bigger & there’ll be way more stuff in engine bay. Main engines actually do majority of work in turning ship vertical before landing.

5.5K likes200 RT97 replies
Aug 17, 2020

@PPathole The “early investor” part of Tesla is wrong. That bs has been pushed by Eberhard (worst person I’ve ever worked with and that’s saying something!) for years.

1.3K likes55 RT94 replies
Aug 16, 2020

@teslaownersSV @PPathole @ICannot_Enough @flcnhvy @Tesla A lot of work remains. Technically, we have it working in sim with FPGAs at ~0.01% capability. This will be a true supercomputer.

1.0K likes56 RT63 replies
Aug 15, 2020

@Erdayastronaut The FSD improvement will come as a quantum leap, because it’s a fundamental architectural rewrite, not an incremental tweak. I drive the bleeding edge alpha build in my car personally. Almost at zero interventions between home & work. Limited public release in 6 to 10 weeks.

8.3K likes845 RT388 replies
Aug 14, 2020

@Teslarati Tesla team is working hard on increasing Powerwall production

4.0K likes144 RT158 replies
Aug 9, 2020

@SamTalksTesla Working on my Mom’s cousin’s farm in Saskatchewan at age 17. I think we did a barn-raising that day.

10.7K likes218 RT279 replies
Aug 2, 2020

@Erdayastronaut @SpaceX However, we did prove that a SuperDraco landing could work by doing a propulsive hover https://t.co/Mr7VFIQwWf https://m.youtube.com/watch?v=07Pm8ZY0XJI

2.6K likes127 RT75 replies
Jul 30, 2020

@keith31345919 @PPathole We’ve been working on this for about 4 years. Originally meant for rabies vaccine, but it can make pretty much anything that’s an RNA sequence. Critical path is completing human trials.

1.7K likes205 RT108 replies
First Principles AI
First Principles AI
Ask anything about Elon
5 free

Ask anything about Elon — companies, predictions, tweets, controversies, vehicles, family.