mbesto

Why on earth is Canva parternering here? This is literally eating their world.

> There's no shame in being homogenous and obvious, though.

The real world analog is this...

The reason people (especially Americans) stay in Marriott property hotels is because they are homogenous. If all I want to do is travel to Phoenix, AZ for work I want to know that the hotel room has the same mattress, desk, TV, customer service, etc. There is real legitimate value to that. So I'll book the Courtyard in Phoenix because I know exactly what I'm going to get.

On the other hand, when I'm traveling the Amalfi Coast in Italy, I want the Airbnb experience. Sure the bed is stiff, there's no A/C, and the 80 year old door frame is hard to close, but there is something magical about it.

And overfitting benchmarks can easily be gamed. Yet here we are with the top HN comment on the HN Mythos thread outlining it's benchmarking performance gains.

I guess we'll never learn.

> That's why we don't recommend purchasing a new machine. Existing machine is no cost for you to run this.

You misunderstood. If the ROI is there, there is enough capital in existence for you to accelerate your profit. So why even deal with the complexity of renting people's hardware when you can do it yourself?

That wasn't my point but okay.

This gets into game theory and doesn't follow a predetermined flow like you're suggesting.

It's a bit like playing poker...does the guy who just went all in have pocket Aces or not?

Correct. This is an important distinction.

Matt Levine has a good take on this. His informal, distilled definition is essentially:

- There is a time gap where insiders know something material and the public does not.

- Someone with access to that material nonpublic information, who is not supposed to use it for personal gain, trades on it anyway.

- That conduct is treated by courts as a deceptive scheme against the less‑informed trading counterparty and against the information’s rightful “owner.”

In other words - you profit at another person's expense (e.g. stealing) because you have information and the other person doesn't.

Two scenarios:

(1) a US Naval Officer knows about a strike 24 hours before it happens and places a bet against someone who doesn't have knowledge about the strike.

(2) Neither a US Naval Officer knows about a strike 24 hours before it happens and someone who doesn't have knowledge about the strike do nothing.

Scenario 1 is (or should be) illegal because the officer is using the information for personal gain, when the information was explicitly given to them for national defense reasons (thus violating the rightful owner clause).

> The bottleneck is no longer engineering. It’s moving up the stack to judgment, customer insight for desired outcomes and distribution.

My posit is this: engineering never was the bottleneck, or at least hasn't been for 10 years now. Frameworks and best practices are pretty well known at this point. AI is simply exposing this reality to engineers' faces.

Proof point - most publicly traded SaaS first businesses S&M equals their R&D spend, if not dwarfs it. You're going to see this even more lopsided going forward.

Why not just Nginx Proxy Manager? Solves both the Proxy issue as well as TLS/SSL.

https://nginxproxymanager.com/

This is reductionist and myopic. I've personally been through building forms online and it's hell to try to find consensus on perhaps the most common forms used online.

Let's take a credit card form:

- Do I let the user copy and paste values in?

- Do I let them use IE6?

- Do I need to test for the user using an esotoric browser (Brave) with an esoteric password manager (KeePassXC)?

- Do I make it accessible for someone's OpenClaw bot to use it?

- Do I make it inaccessible to a nefarious actor who uses OpenClaw to use it?

I could go on...

Balancing accessibility and usability is hard.[0]

[0] Steve Yegge's platform rant - https://gist.github.com/chitchcock/1281611

The way I see it is more like this:

- Skills help the LLM answer the "how" to interact with API/CLIs from your original prompt

- API is what actually sends/receives the interaction/request

- CLI is the actual doing / instruct set of the interaction/request

- MCP helps the LLM understand what is available from the CLI and API

They are all complementary.

The founder of GitButler is the co-founder of GitHub. It doesn't matter what he builds, the VC is going to throw money at them.

Exactly. This is called overfitting and it's most definitely a thing.

> We're gonna need some new benchmarks...

You can't consistently benchmark something that is qualitative by nature. I'm struggling to understand how people don't understand this.

> complexity emerged over time

So complexity is an issue? I don't get it. SFDC is an incredibly complex system that makes billions of dollars. Tell me why I would NOT want to be able to create a system like that with an automated tool?

> Is there any evidence of this being intentional?

The evidence is in the code! If you didn't intend for a capability to be there then why is it in the code?

> if they truly wanted that, they have about 10 better approaches for it, if they don't care about other things.

How so? What other approaches do they have that get this much data with little potential for reputational harm? This is a very common way to create plausible deniability ("we use it for improving our service, we don't know what we'll need so we just take everything and figure it out later") and then just revert the capability when people complain.

Signal?

> All AI is doing is speed running your code base into a legacy system

Are you implying legacy systems stop growing because I didn't mean to imply those companies stop growing.

I've seen enough dirty code (900+ tech diligences over the last 12 years) to know that many businesses are successful in spite of having bad code.

It's a proxy for complexity.

Or a homelab using Proxmox or Unraid.

Uber ran their ridesharing at a loss for years. This is a very common way to gain market share.

But 600 kWh is about ~4k kg, no? Isn't that like the max hauling a sprinter can do? So doesn't this just get you a bunch of range at the "cost" of not being able to haul anything or am I missing something?

I don't get why I would use Claude Code when OpenCode, Cursor, Zed, etc. all exist, are "free" and work with virtually any llm. Seems like a weird use case unless I'm missing something.

> It's widely understood that the big players are making profit on inference.

This is most definitely not widely understood. We still don't know yet. There's tons of discussions about people disagreeing on whether it really is profitable. Unless you have proof, don't say "this is widely understood".

Wait, you put 600kWh of batteries on a sprinter van?!? I wanna know more...that's insane (base eSprinter is like ~115 kWh, right?)

This is a fair point as it's not just simply using ethanol for gasoline. This article goes into more depth about it: https://www.pnas.org/doi/10.1073/pnas.2501605122

There's lot of factors at play here:

- Location for generating PV

- Redistribution of food (both for livestock and human) production

- Environmental impacts of PV vs livestock vs depletion of native prairies

Point still stands...if you replaced all of the land used to produce ethanol with PV, you would create a surplus of energy that is higher than anything we could imaginably consume today (hint - China is essentially already doing this)

We already produce enough food. Rooftop solar by definition is an inefficient use of resources.

> why it is not done yet?

Whoa lots to unpack here. I'll summarize:

- It is already happening to some extent (it's cheaper)

- Try explaining to farmers to do away with their livelihood and retrain them to running a solar farm

- Entrenched bureaucracy and gov subsidies

Fun fact, 12 million hectares of land of used to produce corn used for ethanol which is used to produce gas. I'll let you draw the conclusion.

https://news.cornell.edu/stories/2025/04/trading-some-corn-e...