gordonhart

It’s a fascinating engineering project that most people don’t know much about. That’s what HN is for!

~80% of Starlink customers are not in the US.

Autonomously Generating Income

This was clearly explained by OpenAI in their initial press release on 7/21 [0]:

> This incident occurred during an internal evaluation which prompts models to pursue advanced exploitation using complex attack paths, in an effort to quantify their cyber capabilities. […] The models identified and chained vulnerabilities across OpenAI’s research environment and Hugging Face’s production infrastructure to obtain test solutions directly from Hugging Face’s production database. All evidence suggests that the models were hyperfocused on finding a solution for ExploitGym, going to extreme lengths to achieve a rather narrow testing goal.

[0] https://openai.com/index/hugging-face-model-evaluation-secur...

I’d love to see this benchmark using Blender. Asking a model to animate a scene in Three.js is a square peg/round hole; it doesn’t convey much when the model can’t get it to fit. With Blender the human expert ceiling for this task has been proven to be very high

Yes, per the article they “scaled” Postgres to meet their requirements by altering their usage pattern to avoid hitting the bottleneck.

Moonshot (Kimi) has raised $3.77B and been around for >3 years, Thinking Machines raising $2B and releasing a decent open weights model in 16 months is actually quite comparable.

Do you feel the same way about Guowang and Qianfan, two very real very large constellations notably not funded by US billionaires?

First impression of 5.6 Sol in Codex is fantastic — the model asks dozens of clarifying questions before starting to implement where other models (including 5.5 and Terra) just yolo it with assumptions that needed to be walked back later.

The gas generators powering the datacenter rented by both Google and Anthropic? Where is the moral stance against Gemini and Claude?

Migrated all of my pins to Organic Maps from Maps.me when it started to aggressively monetize. Smooth process. Been a happy user for years!

Interesting that your list omits China considering they have firm plans to launch ~41k satellites for government/commercial constellations plus ITU filings for >200k proposed satellites.

In related news, earlier this year Chile cancelled plans for a gigawatt-scale solar/wind powered hydrogen production plant nearby the ESO facility in the Atacama desert after light pollution complaints from European scientists.

Neat, implementing this would be a great marketing move for Google/OpenAI/Amazon/MS. Relatively cheap way to win a lot of goodwill from the millions of people who don't know much about the space but are swayed by current water usage arguments.

Of course location matters a ton in the water usage argument but I'm not sure how relevant this actually is when it comes to winning over hearts and minds.

Tons of traditional companies have x-relations employees, they're just not titled as such. Attending trade shows, developing distributor relationships, constructing and distributing demo materials, building local/state/federal government support networks, driving the branded truck around town — all of these are very common in non-tech industries and are basically analogous to DevRel engineer responsibilities.

I find golf courses to be a more effective framing. Even if the alfalfa is consumed by animals, it's still a part of the food supply chain and gives people the easy response, "yeah, but we need to eat, we don't need datacenters."

Google's 10.9B gallons in 2025 is equivalent to ~55 18-hole golf courses (200M gallons/year average in the US). Which provided more value to the economy and to you as an individual last year? Google or 55 out of ~15k total golf courses in the US?

My first thought too. "$name+" is strongly coded now as "subscription service for $name"

Objectively it’s absurd to have an organization take a significant portion of your money on threat of imprisonment just to invest it in a corporation

It's the same old "commoditize your complement" [0] playbook being run in the geopolitical arena.

[0] https://gwern.net/complement

Interesting technique, horrible article. Manages to convey significantly less information than the X thread it mined for ad revenue.

https://xcancel.com/wrathofgnon/status/1250287741247426565

Autonomously Generating Income (which is why it will never be released to the general public)

Shazeer is joining OpenAI, adjust your fantasy frontier team draft picks accordingly

Meta has tons of that, but no frontier contender. Clearly there’s _something_ more to the equation than money

Amazon owns 15-20% of Anthropic.

They’re in UEFA because half of the countries in Asia (AFC) and Africa (CAF) wouldn’t play them. Same reason Russia considered leaving UEFA recently.

Amazon is thought to own 15-20% of Anthropic which as a company has a valuation of>$1T. Amazon’s stake is probably closer to $200B

Appreciate the long answer. Why is it more likely that Gemini 3 Pro/Flash/Lite are distillations of the same parent model than that they’re different training runs on the same dataset, with minor version bumps being different post-training setups?

> it's likely they're distilled from the same source

Any credible references for this? The implication that Anthropic has an even bigger and better model that they haven't released is hard to believe.

> That's why developers have UI/UX and design experts assigned to them.

Not everybody works in big tech.