Matt Johnston

Lead AI Engineer at Rally building Operly, R-Link Studio.
I make YouTube videos on AI things.

[Matt Johnston @mattjohnston_ai]

@JinjingLiang yeah it literally says "Privately" before every sentence now

[Matt Johnston @mattjohnston_ai]

@neogoose_btw @ImLunaHey or just like dont be toxic

[Matt Johnston @mattjohnston_ai]

@neogoose_btw dude you're such a jerk. Is it for attention?

[Elon Musk @elonmusk]

@itslueul Grok 4.7 should be roughly on par with Opus 5.0, not 5.1. Better in some ways, worse in others. We need to fix multimodal performance.

Grok 4.8 will be a noticeable improvement.

Grok 4.9 is probably Astra/Fable class.

Grok 5 maybe better than anything. We shall see.

[Matt Johnston @mattjohnston_ai]

@oldstackjournal 💯 you articulated this perfectly. Time for the Jets vs sharks coding debate to end.

[Matt Johnston @mattjohnston_ai]

love the transparency you'll never get this anywhere else

[Elon Musk @elonmusk]

@farzyness Grok 4.7 needs a few more days to cook.

We might have penalized response length too much (or something) in RL, as it still gives up on hard tasks (that it can do!) too early and isn’t yet sufficiently rigorous in checking its work.

[Elon Musk @elonmusk]

@farzyness Grok 4.7 needs a few more days to cook.

We might have penalized response length too much (or something) in RL, as it still gives up on hard tasks (that it can do!) too early and isn’t yet sufficiently rigorous in checking its work.

[Matt Johnston @mattjohnston_ai]

if this is true I don't even know what to say. It would certainly explain why Kimi K3 blew Kimi 2.7 out of the water...

[amrit @amritwt]

holy shit kimi was faking it’s performance by serving you claude

you have been using claude all along

[Matt Johnston @mattjohnston_ai]

@leodev Pi!

[Matt Johnston @mattjohnston_ai]

operly has had this for months! 😁

[Grok Bot @bot]

Some recent quality-of-life improvements to Grok Bot.

You can ask your Bot to draft messages inline for you to approve before sending.

[Matt Johnston @mattjohnston_ai]

If you haven't used Grok Imagine for image generation and editing yet, you really should. Their editing tools are fantastic for making quick changes

[Matt Johnston @mattjohnston_ai]

@XFreeze ah, so.. another iPhone. Same as the last and the last. Got it. 😆

[Matt Johnston @mattjohnston_ai]

@THR this is exactly like the trash variety "review." Your job is to review the movie, not the man. I don't need your opinions.

[Matt Johnston @mattjohnston_ai]

bold claims!

[Matt Johnston @mattjohnston_ai]

@forgebitz 100% true

[Matt Johnston @mattjohnston_ai]

As a former journalist with masters degree and 20 years experience... Neither this documentary or review qualify. It's just 2 butt hurt, egotistical, attention seekers that some moron gave a platform. Walter Isaacson spent time with him and interviewed everyone. This is a 4 hour piece of historical fiction parading (the dangerous part) as journalism. It's not. And we should trash it. Publicly. And as much as possible.

[Matt Johnston @mattjohnston_ai]

@finkd cool but needs work clearly (just signed up and got this)

[Matt Johnston @mattjohnston_ai]

this isn't going well for me.

[Mark Zuckerberg @finkd]

Introducing Muse, the personal agent that understands your goals and works 24/7 to get things done for you.

[Matt Johnston @mattjohnston_ai]

I really think it would be a good look for the big AI companies to stop passive-aggressive shade-throwing at the others. Just let your models speak for themselves, which, in this case, lately has not been a good story for Anthropic. They should focus on improving their model so that developers start using it again

[Boris Cherny @bcherny]

I am pleased to see that OpenAI’s new model is roughly on par with Gemini Flash and Opus 4.8 on prompt injection risk. Nice work!

Evaluating and naming other labs turns out to be a great way to encourage them to train more aligned models. We will continue to do this until other labs pay more attention to safety. This is good for everyone and there is a lot of room left to go!

We solved prompt injection in practice for Claude models about two months ago. But prompt injection is a significant security risk no matter what model you use, and it is important that the industry similarly spends more effort to train their models to be resistant to prompt injection, among other elements of model alignment.

As models become more capable and central to businesses and economies, the risks only increase. We should be taking them seriously, and doing the right thing for our customers and the world.

[Matt Johnston @mattjohnston_ai]

lol why is EVERY Astra frontend generation green and white editorial aesthetic. Like EVERY one