Landon Garrison

Creator of Tutory | Post random thoughts about AI

[Landon Garrison @lando_garrison]

@apples_jimmy In 6 months time, we might be in a situation where frontier models are considered advanced weapons technology and is treated as such, slowly building towards bigger nationalization of the labs. AI is immensely unpopular with voters in the US, so this will be widely supported.

[Landon Garrison @lando_garrison]

@imjustnewatai They really almost need to sandbag this launch or at minimum significantly underplay the capabilities it seems. If they are at or above Mythos, are they getting export controlled?

[Jimmy Apples 🍎/acc @apples_jimmy]

@AnthropicAI I mean it’s what Dario’s been asking for

[Landon Garrison @lando_garrison]

Nationalization inbound. This is such a disappointing direction we are headed in.

At the same time, Dario, Sam, Elon and the rest deserve this. They fear mongered for far too long and now they will get called on it.

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[Landon Garrison @lando_garrison]

And they will win over more customers. I’m glad to see the shift slowly turn back to who has the best model that is also cost efficient.

Pushing the frontier is amazing, don’t get me wrong. But if all we have is expensive models that most people can’t use, it defeats the purpose

[The Wall Street Journal @WSJ]

OpenAI is considering drastic price cuts as it seeks to win over customers from archrival Anthropic https://t.co/vwWcJJd1x8

[Landon Garrison @lando_garrison]

https://t.co/3MfNkLhMRh

I’m surprised more people are not talking about this in lieu of president Trump expressing interest in nationalizing the labs, or at least taking significant steaks in the labs.

General consumer AI is on limited time.

[Landon Garrison @lando_garrison]

Here is the beginning of nationalized ai. Was fun while it lasted lol

https://t.co/KMBS6v853H

[Landon Garrison @lando_garrison]

@JustJerry121 The agent does commit the memories automatically, but you can absolutely see the difference if you wanted too. At the end of the day, it is just a collection of md files. That’s really good feedback though!

[Landon Garrison @lando_garrison]

Built a terminal-based coding agent called Kin. The goal wasn't better code generation, it was better context. I want an agent that understands my project, challenges ideas, makes suggestions, and stops me from repeating myself. https://t.co/B4EHj5bjnM

[Landon Garrison @lando_garrison]

@JustJerry121 It stores memory files, project notes and session history. One key thing is Kin also has a built in reflection and wake cycle. So everyday (or night), Kin can reflect on what you talked about and worked on that day to improve and brainstorm new things to work on!

[Landon Garrison @lando_garrison]

This is the type of shit that makes DeepSeek’s/Kimi/Minimax’s popularity explode. If we’re in for continual price increases, we’re basically forced to look elsewhere for cheaper inference.

[Enrico - big-AGI @enricoros]

Disappointing pricing trend with Gemini 3.5 Flash.

22.5x pricier than 2.0 Flash which came out 15 months ago ($9.00 vs $0.40).

Are Flash models supposed to get this much more expensive, or is Pro just being renamed to Flash?

[Landon Garrison @lando_garrison]

@scaling01 Was my exact reaction when I went to AA because I was curious if Google had made strides in token efficiency.

The model is good to use so far, I can’t honestly say it’s bad. My gripe right now is “ is this 3x price increase good?” Jury is out but my early usage is probably not.

[Landon Garrison @lando_garrison]

@sama When are we getting the next voice upgrade?

[Landon Garrison @lando_garrison]

@riadbenguella On tasks where they are small to medium in length, relatively straight forward complexity, both are about the same, maybe Claude code slightly.

When the task blows up or has a lot of nuance/tricky complexity, Codex and it’s not even close.

[Landon Garrison @lando_garrison]

In hindsight, this was just the beginning. By the end of the year we will have basically seen a doubling of pricing across the industry.

Wild

[Landon Garrison @lando_garrison]

Observation: Google and OpenAI has raised prices at least once.

OpenAI —> GPT-5.2
Google —> Gemini 2.5 to 3 series

But Anthropic was the only one to drop prices this year or not raise them.

I wonder if this is a sign that prices on the frontier will increase like Dario said.

[Landon Garrison @lando_garrison]

@scaling01 This is what I feel like labs get wrong often. I will be a lot more forgiving and willing to give the model a fair assessment if it’s fairly priced.

[Landon Garrison @lando_garrison]

And with that, I also have to call out DeepSeek V4. I’m not going to lie I kinda thought DeepSeek was largely finished with being on the frontier. Couldn’t have been more wrong.

This thing, like GPT-5.5 feels good to use and has more depth than before.

[Landon Garrison @lando_garrison]

GPT-5.5 is an unreal model. Nuance is amazing and just has a very “elegant” feel to it and a very high effort model.

My only critique is pricing. I hate that we’ve had 3 consecutive model releases that are increasing costs to the point of being more expensive than opus.

[Landon Garrison @lando_garrison]

@apples_jimmy I think the bigger question is was it trained on Huawei chips or not?

If yes that’s a much bigger story

[Landon Garrison @lando_garrison]

This to me feels alot like when o1 first came out. The graphs are so starkly increased in some areas that it’s hard to grasp how much better this model is.

[Anthropic @AnthropicAI]

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.

It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.
https://t.co/NQ7IfEtYk7