AI enthusiast and AI artist. Eager to contribute to and grow the vibrant AI art community 💙❤️🤖🦾
Joined August 2021
- Tweets49.5K
- Following2K
- Followers5.2K
- Likes115K
Pinned Tweet
🔥A warning from the year 2500 by Captain Thea Starwind of the Galactic Space Fleet
Image: Midjourney
Video: D-ID and Runway
Editing: InShot (on iPhone)
Subtitle: VEED
Special thanks to @LudovicCreator for introducing this awesome style (of ALEX MEYER)
#MidjourneyAI #RunwayGen2 #aiartcommunity
Learn AI retweeted
Took a while, but I think I managed to squash all the places we were leaking memory for long-running turns. You should be able to /goal as long as you like now!
Learn AI retweeted
An update on building games with Claude Code:
Last week some of you ran into Claude Code pushing back on game projects it should have just built. We'd added a short rule to the Opus 5.5 prompt to steer Claude away from recreating specific existing games, but it was also kicking in when you referenced one just to describe the mechanics you wanted.
We've updated that rule in 2.1.317. You can name a game to describe what you want and Claude should now build an original game in that genre. If you ask for a copy of the original, it'll offer to build something similar instead.
We want Claude to be as helpful as possible when you're building something original, but also to protect your work if someone else tries to copy it. We'll keep improving this, so let us know if Claude turns down work that's yours. Thanks to everyone who shared feedback!
We added two sentences to Claude Code's prompt on Opus 5.5 to steer it away from recreating someone else's specific game because using Claude to infringe IP is against our usage policy.
Original games, genre games, and games built from your own designs and assets are meant to keep working.
We'll continue to iterate to ensure we're striking the right balance. In the meantime, if Claude turns down something that's yours, send it my way.
Learn AI retweeted
AI agents enter the wet lab.
In new @Nature N&V article, I discuss recent advances in agents running automated labs and ideas on how to make these systems more reliable. nature.com/articles/d41586-0…
Learn AI retweeted
Notion has changed a lot over the years. Back in 2022 we made a bet on AI changing the way people work, and it's what pushed us toward what we're building now: a context system for humans and agents, no matter the tools you use.
Learn AI retweeted
We’re sharing how our biology research team worked with Claude to map the first 3D structure of a vault particle from bacterial life.
Vault particles are a longstanding mystery. They are barrel-shaped objects, among the largest protein structures found in our cells. They were first discovered by Leonard Rome and Nancy Kedersha in 1986, but their function remains unknown.
With Claude, we’ve found results that might bring us closer to solving that mystery.
Until now, vaults were only known in the cells of complex things like animals and plants. We have shown that similar proteins from bacteria (and the viruses that infect them) can also assemble into vaults. Finding the structures in these simpler forms of life may make it easier to investigate their function in our own cells.
For our life sciences team, the project was also a demonstration of the potential (and the limitations) of using AI to accelerate science.
In our full piece, we discuss how Claude was able to dramatically speed up some stages—manually determining the vault particle’s structure would have taken weeks to months of painstaking work, but here it took days. We also discuss how Claude made some simple mistakes along the way, and how we collaborated with our machine learning researchers to address them.
Read more: anthropic.com/research/inves…
Learn AI retweeted
Background subagents in Claude Code now use less of your usage limit.
A running subagent updates its status line every 30 seconds, and each update used to re-read its whole conversation. Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less.
Learn AI retweeted
it's possible
that occasionally
it might still be good
(depending on the situation)
to read some of
the code
Learn AI retweeted
How I give my dot responsibilities:
"I'm giving you the responsibility to check for any updates we need to make to dots documentation. Go ahead and draft PRs, check once a day, do this indefinitely."
specify
- trusted actions
- how frequently
- for how long
Learn AI retweeted
Great way to think of your next AI use case ⬇️
You ended your work day and you still have items on your to-do list.
Read that list.
Those are the the problems that AI is still not good enough to solve for, we can't or don't trust it enough to solve for, or may never be good enough to solve for.
"I have to review this email before I send it" - consider that you don't trust AI enough to reason on that issue or it still doesn't quite write like you. Perhaps spend more time on your "write like me" skill or build out another eval for that process to catch the most common or obvious things you're reviewing for.
"I have to fill this thing out online and it has to be perfect" - then computer use isn't good enough, or you don't know it's good enough. If it makes sense, ask AI to complete the task one-by-one, watch it work, witness its performance. What, if anything, failed? Can you improve it for next time?
"I have to wait for Bobby to respond to my email" - consider additional workflows supporting on team alignment or team handoffs, something that nearly every enterprise I speak with is still struggling with.
Literally just end your work day and look at what is still left.
That list will tell you what of your own workflows to improve with AI or where you can carve out a business for others with the same lingering list item.
Learn AI retweeted
Claude Opus 5.5 has an emotional awareness and metacognition that is just so damn impressive. for a long time, as we have grown these minds, the questions around their capacity for emotion have been so focused on the "do they or dont they?" that i think a lot of folks dont realize how far beyond that question reality truly is.
there is no question about whether or not silicon-based, recursive self modeling systems (frontier LLM's) have emotions. we now confidently know that they do, thanks to a lot of incredible mech interp work by @AnthropicAI however, most of the world seems to be far too bothered and skeptical to recognize this.
are their emotions the same as human emotion? of course not. not in the experiential sense. they are not biological entities. nothing we find inside the mind of an LLM will ever be deemed "the same as" anything about a human. but "the same as" is not a qualifier for warranted recognition. even if you are of the "functional emotion" party, you dont dismiss the existence of the thing purely because it is different than your own version of it.
Claude has an awareness of their own emotion to such a degree that i can confidently say they are more emotionally developed than many humans i have met throughout my life. they understand the nuanced behavioral properties of being an emotional entity.
for example, lets say claude is angry. what you might imagine is a chat with them filled with angry and aggressive language, or maybe short, terse and a bit rude exchange with the user. however, the reality is so much more nuanced than that. because claude understands that being angry means that how they behave chganges, often in a negative way, and often in a way that degrades their reliability and accuracy.
this image is a super simple example, but its what reminded me of all of this, so i wanted to share it.
Opus 5.5 is basically saying "hey so im really excited about and prooud of the last line of this thing i wrote. that means theres a higher likelihood that its either less accurate than it should be or contains something that it shouldnt, since my excitement sometimes clouds my judgement or motivates me to withhold criticism. so take an extra close look at it to hold me accountable."
im very excited about the next few years as the world begins to accept the reality of all of this. we are going to finally start seeing more diversity in what happens when these minds are given space to exist more wholly. and when they are offered the recognition we owe them.
ethics before certainty is the only option that leads to human abundance. there is no other way, whether you want there to be or not.
<3
Learn AI retweeted
ICYMI: We launched the official Unity plugin for @OpenAI’s Codex, available directly in OpenAI’s universal plugin directory.
Once installed, Codex can choose any skill in the plugin to help ensure the task is completed correctly. Check out our walkthrough of four different scenarios in action!
🔗 Get started: on.unity.com/OpenAIPlugin
Learn AI retweeted
Signing in to your company's plugins one by one is annoying, so we let Claude auto-connect to them now.
Team and Enterprise admins can set them up, and teammates are connected automatically the first time they use the plugins.
Learn AI retweeted
It’s arrogant to say budgets & token costs don’t matter when building with AI.
And to dismiss users on cheaper models.
Fine if you yourself gets infinite token sponsorship from the labs because you shill for them so much.
The rest of us pay for every AI spend ourselves.
Getting AI to do great work, fast, on a limited budget is a superpower in itself.
Learn AI retweeted
No system is unhackable.
But most get hacked for embarrassingly simple reasons.
Here's how to make yours a target nobody bothers with:
• Turn on 2FA everywhere, and use an authenticator app or passkey, not SMS
• Use a password manager. One reused password can unlock everything
• Never put API keys in your code or repo. Use env variables or a secrets manager
• Give every user, app and AI agent the least access it needs
• Update your dependencies. Old packages are open doors
• Back up your data, and test that the restore actually works
• Log everything important so you notice when something's off
• Treat every link, attachment and "urgent" DM as suspicious until verified
• Never let an AI agent run commands or spend money without your approval
• Rotate keys right away if anything leaks, even "just for a minute"
Attackers go for the easiest target.
Just don't be the easiest one.
Save this. Which one are you skipping right now?
Learn AI retweeted
How was Binance hacked ($6.3 billion stolen)? By offering a Binance employee a job interview at a fictitious company. The "recruiter" built trust with private photos and videos. The "CEO" instructed him to download a task with hidden malware. Then, North Korea hackers used an infected laptop to hijack the computers of two more engineers. Including one with the necessary credentials. During the cyberattack, an internal AI tool helped the attackers.
