w1

Hasura has been a godsend for me. I'll be using it for all my MVPs for the foreseeable future.

Super cool! Also, a character visualizes data this way in Watts' "Blindsight" novel.

It goes faster than the speed of light _in water_. While the speed of light in a vacuum is fixed, light will pass slower through different mediums (i.e. it goes slower in water than air, which is how you get refraction).

It is very extensive. It has a lot of fairly domain-specific work.

You have to let them come in, but they have to pay you to lease the surface easement they use from you. That lease doesn't begin to approach the value of the minerals though.

Thank you so much! I'm a ML/scientific computing guy exploring more generic web design, and this was my first crack at it.

N-body problem gravity simulator, in vanilla js:

https://www.thorbjorn444.com

I did first try to setup email, but I couldn't figure out how to do that without providing a static ip address. This was a "what can I make work in the two hours before we drive across the country" solution lol.

It looks like they are, thanks! This is just what I could hack together before I went on Christmas break a couple years back lol.

Lol I may or may not currently be using a Twitter account and cron to broadcast any changes to my (non-static) home ip address, encrypted, so I can ssh into my workstation when I'm on vacation.

I've been working in this space for a few years, as an engineer and now founder, and I think a lot of savvy investors are beyond this.

In my experience, everyone who is informed acknowledges that ML is very powerful, but the algorithms are widely accessible.

Instead, it seems that investors are looking for a company that protects itself with a proprietary source of data that allows for results that are unobtainable by competitors. Also, to a lesser extent, domain expertise that allows them to tailor existing ML architectures specifically for the problem at hand.

As a Python developer moving into web development, I've really enjoyed using Plotly's Dash.

Usually, one can, with no ill-effects. Many operators run their marginally economic wells for only a few hours per day already.

A recent(ly featured on the front page) example where synthetic data is helping solve an unsolved problem: https://eng.uber.com/generative-teaching-networks/

I'm being pedantic by posting this though, because their approach involved learning the synthetic data generator simultaneously with the classifier/whatever that you're training. That is not relevant for a static synthetic data source.

The issue isn't that they found a copy of the test data online (The test input data was provided to them as part of the problem.)

The issue is that they manually labeled the test data, and then pretended they didn't.

The competition objective is to provide an ML solution that produces labels for the test data, showing your work with code (to prove you didn't just hand label the data.)

Instead, they did manually label the data, and hid their manual labels in the id column of that external data source.

The look like a fancy version of a gauging tool [1]. Each hole could have been used to measure the diameter of a different sized rod, sphere, or coin..

[1] https://www.eastwood.com/sheet-metal-gauge.html

Annulus is the space between the outer and inner pipe.

Most new pipelines are inspected currently with "drones", called smart pipeline inspection gauges, or smart PIGs, that use ultrasonic or magnetic flux to measure the wall thickness over the entire length of the pipe [1]. The PIGs are propelled through the pipe via the working fluid in the pipeline.

I think creating a similar device to travel in the annulus of cased pipe would be much more difficult. The centralizers used to keep the inner pipe centered in the outer pipe would pose a pretty big obstacle. Plus, there would be nothing propelling it forward, and communicating with the device through a steel pipe would be very difficult.

[1] https://primis.phmsa.dot.gov/comm/FactSheets/FSSmartPig.htm?... [2] https://kwikzip.com/ (example)

They used to do this, before modern coating techniques. The issue is that if any moisture gets inside that annulus (which will happen eventually), you get corrosion. It's also then very hard to inspect the integrity of that inner pipe with traditional techniques, like guided ultrasonics.

ironically enough, there are more and more electric drilling rigs now, especially when drilling in-fill wells in established fields or near/in cities. they don't use battery power though, just temporary power lines and a mobile transformer skid.

Per DeepMind's blog post[1], an agent was initially trained via supervised learning on pro matches. Then, the agent was forked repeatedly, as the population of agents learned via tournament-style self-play. So, while initial strategies could have been seeded by pro play styles, the final models were the result of models learning from games with other models.

[1] https://deepmind.com/blog/alphastar-mastering-real-time-stra...

This. I pivoted by finding applications of ML in my day-to-day engineering role, until I got enough internal traction to justify a title change from "____ Engineer" to "Data Scientist." About a year later, I jumped ship to a startup applying ML to my domain, and I am contemplating making the move to FAANG next.

This may sound odd, but my biggest improvement in ability to have casual conversation with anybody came after a few months of beginner improv classes (at Second City). I'm not sure if it was being forced to quickly say _something_, or just being made more comfortable in awkward situations, but those classes did wonders for my long sought-after transformation from "weird conversationalist" to "normal guy".

Thank you, being reasonable is really important, and being unreasonable when investing can be very expensive!

Producing 55k Model 3's per quarter [1] will definitely help with the cash flow situation. I think the biggest questions that determine if they can be self-sustaining are:

- Can they ramp up Model 3 production quickly enough to meet the debt obligations they have accumulated to date?

- What will the average Model 3 margin be, given the price distribution from 60k to 35k?

The bull case [2] aligns with your basic assumptions that the 3 will add significant cash flow that allows them to get over the "debt maturity hump". Note that their analysis does include other sources of cash flow (energy credits, remaining line of credit).

The bear case, basically, disagrees. They suppose that there will not be sufficient demand for the higher end Model 3, given increased competition from vehicles like 2019 Jaguar I-Pace (69.5k, 240mi); Chevy Bolt (37.5k, 235mi); and maybe Hyundai Kona EV (? $, 250 mi). This could then put Tesla into a negative spiral, where shrinking orders cause their accounts payable to not keep up with their lagged accounts receivable.

Also it's worth noting that they are probably on track to spend 2.5B in capital on PP&E this year based on 1.25B in first six months [3], and spent 3.6B, 1.3B, and 1.6B in the previous three years[4]. Adding in another 0.75 - 1.0 in selling/general/admin and R&D costs per quarter, on top of the cost of producing their cars and servicing their debt, makes even the 3B/yr in Model 3 gross profit case less of a slam dunk for Tesla success.

I think it really could go either way at this point, and the next six months will be telling. Elon does has a tendency to just make things work, but I am taking a risk here and siding more with the negative case.

[1] https://www.bloomberg.com/graphics/2018-tesla-tracker/

[2] https://www.cnbc.com/2018/10/09/tesla-on-path-to-profitabili...

[3] http://ir.tesla.com/node/18946/html

[4] http://ir.tesla.com/node/18501/html

Those are really good questions. I don't know what form a transfer of cash from Elon to Tesla like that would take. Tesla issuing new shares for him to buy?

Honestly, on paper there isn't any reason why Tesla can't just sell additional shares to the market in general, not necessarily just to Elon, in order to get enough money to cover their upcoming debt payments. Elon has publicly said he won't sell shares because Tesla won't need the money because it will be profitable[1], but some people are skeptical [2]. Maybe he just doesn't want to walk back his statement?

Also as previously mentioned, he can't sell all of his Tesla stock without paying back some of his personal loans.

[1] https://www.cnbc.com/2018/08/01/musk-says-tesla-wont-be-sell...

[2] https://www.businessinsider.com/teslas-upcoming-debt-payment...

Just did some rough math on this. He's reported to have 55% of SpaceX, recently valued at 27.5B [1], and ~19% of Tesla [2], most recently at ~44B, for a combined net worth of ~23B (before any debt). He therefore definitely couldn't personally buy out Tesla at current prices, and any significant investment of personal funds would require the sale of SpaceX shares.

Also, much of his liquid pocket money has come from loans backed by his Tesla Stock. This could be a problem if Tesla stock price drops too much, as the stock he has pledge is required to be no less than a multiple of the money he has borrowed. [3]

Disclosure: I am short Tesla.

[1] https://www.cnbc.com/2018/04/13/equidate-spacex-27-billion-v... [2] https://money.usnews.com/investing/stock-market-news/article... [3] https://www.forbes.com/sites/jimcollins/2018/05/18/musk-has-...

Not really a book, but...

The Berkshire-Hathaway shareholder letters are very entertaining and informative, and reading Warren is probably the best way to deeply understand the last ~40 years of US business climate.

http://www.berkshirehathaway.com/letters/letters.html

The Order of Time spends some time qualitatively discussing quantum gravity, and the audiobook narration by Benedict Cumberbatch is really excellent.

One energy storage method that's relatively cheap, efficient, and scalable is pumped-storage hydroelectricity. There's a 3 GW plant in Virginia that's been in operation since the 80s[1]! There is ultimately minimal need for FF peakers, or large rare-earth batteries, if governments can execute projects like this.

[1] https://en.wikipedia.org/wiki/Bath_County_Pumped_Storage_Sta...

Henry George was right all along! It's amazing to see how similar income levels are after taking rent into account.

This project is still using crowd-sourced sketches though. It's automating the object detection and localization, then placing a sketch of that object in that location, selected at random from the Google QuickDraw dataset.

https://github.com/googlecreativelab/quickdraw-dataset