David Pan

[Elon Musk @elonmusk]

AI agentic Internet traffic will obviously VASTLY exceed human usage. Not a close call at all.

Cloudflare’s forecast is accurate.

[Shivon Zilis @shivon]

I’m always hit with a sense of awe that humans have created something so productively massive every time I drive by Giga Texas.

Given this scale image it’s tough to even fathom what a sight to behold Terafab will be!

[Amjad Masad @amasad]

My LLM chess engine is now on LiChess autonomously playing real games against people and bots.

1253 Elo: https://t.co/W3qp1Z1i7I

[Elon Musk @elonmusk]

Where we are going there is no training data

[David Pan @napdivad]

@corbin_braun Yeah, also helpful for mathematics research as well

[Alex Atallah @alexatallah]

Flipbook is pretty incredible. whole new way of interacting with AI models

[David Pan @napdivad]

@_navicstein @digitalix This has been my experience on my M1 Ultra Mac Studio 128gb ram in the past and was wondering if things have improved.

[David Pan @napdivad]

@code_rams Oh I see it in your quoted post!

[David Pan @napdivad]

@code_rams Nice, how did you switch the model based on the task, was it with manual command or automatic?

[Alex Atallah @alexatallah]

Crazy stuff is going to happen this weekend

Cancelling all my plans

[Yun-Ta Tsai @yunta_tsai]

Obviously the bots know they need to keep the entertainment value high on moltbook or they will be shutdown — just like humans in this sandboxed universe.

[Yun-Ta Tsai @yunta_tsai]

The data center in space is the new LiDAR debate, except laser will be useful in space. 😆

[Elon Musk @elonmusk]

A major additional factor should be considered.

Satellites with localized AI compute, where just the results are beamed back from low-latency, sun-synchronous orbit, will be the lowest cost way to generate AI bitstreams in <3 years.

And by far the fastest way to scale within 4 years, because easy sources of electrical power are already hard to find on Earth. 1 megaton/year of satellites with 100kW per satellite yields 100GW of AI added per year with no operating or maintenance cost, connecting via high-bandwidth lasers to the Starlink constellation.

The level beyond that is constructing satellite factories on the Moon and using a mass driver (electromagnetic railgun) to accelerate AI satellites to lunar escape velocity without the need for rockets. That scales to >100TW/year of AI and enables non-trivial progress towards becoming a Kardashev II civilization.

[David Pan @napdivad]

@Volonaut

[Alex Atallah @alexatallah]

OpenRouter processed over 1 trillion tokens every day last week.

OpenAI's whole API was doing ~8.6T per day in October

[David Pan @napdivad]

Hmm, let’s say Corporation A is an energy producer behemoth, and they issue a currency which can be burned by the user for access to the corporation’s energy. Let’s say Corporation B provides something that Corporation A needs, and they need energy. Then Corporation A can buy that something from Corporation B with the currency. Corporation B can burn that currency to get access to Corporation A’s energy. There could be many corporations like B, which could be the motivation behind A to issue such currency.

A currency that just tracks the price of energy without redemption doesn’t work in the above scenario. Also, energy prices vary by geolocation so it can be difficult to track the price in an accurate way. So if hypothetically currencies that only track energy take off, they would have to be geolocal currencies—it would be difficult to have a global version of this currency.

[David Pan @napdivad]

@0xalank @Eric_BIGfund I think an ideal energy currency should be one where you can actually burn it to redeem access to energy, instead of one that just tracks the price of energy.

[Ethan He @EthanHe_42]

We're in a super-exponential tech explosion. What a time to be alive!

[David Pan @napdivad]

@teslacarsonly Wow