
Code @Google Creative Lab making AI accessible to everyone. Generative artist finding joy in programs making unexpected pretty pictures
Brooklyn, NY
Joined October 2008
- Tweets4.5K
- Following1.3K
- Followers3.1K
- Likes6.5K
Kyle Phillips retweeted
The 90s called. They said to tell @isofluff they love them.
This was made by @isofluff using a @Raspberry_Pi + an old TV he found here at the office. The intro graphics were all coded by Gemini 4 Argon.
Kyle Phillips retweeted
It's been fun playing with Gemini 4 Argon! We made this video showing lots of our experiments (and my cat 🐱!) Team effort w/ @kkwalker + many friends here.
Nothing like wildfires 1,000 miles away to remind you that we are all connected. ❤️ to Ontario and my old home state Minnesota.
Replying to @it_unprofession
Excellent job leveraging negative effort for maximum strategic impact
this will be the lore spreading through stand-ups for weeks to come
I think Andrew has been consistent in his statements that this work is incredibly personal to him, this is his life's work and the only benchmark he is concerned with is the quality of the language design. If a major company's choices felt like a blight on what you cared most about it might also feel personal and hopefully you would also show this much restraint.
There's only like 5 full-time people making this language.
Replying to @schrockn @jarredsumner
I don't see the problem, they obviously just have very different goals and don't see eye to eye. Neither one seems to say anything wrong, incongruent with other comments or classless. Zig succeeding is good for everyone, as is rust and I like bun no matter how it compiles.
Replying to @addyosmani
the cognitive debt problem problem is real. I use a red-green-refactor system inspired by Googles Conductor extension so I have verifiable breadcrumbs.
I think there's an interesting next step with these artifacts and Anthropic's recent post related to HTML > markdown plans.
For years I have left myself these TODOs in this same note system every day, my behavior hasn't changed, but now I can just wake up, review the work that has been done and explore more ideas.
This was inpsired by @addyosmani piece on the Outer loop xitter.cf/addyosmani/status/2074…
Closing the loop: Once I merge the PR, the agent automatically goes back into my personal notes and marks the TODO as DONE.
The Evidence: it opens a draft pull request on github. If the change is complex, a separate code-reviewer agent leaves a critique as a PR comment.
The Verdict: this is where I sit. I review the PR, inspect the generated spec/plan artifacts and decide whether to merge.
The agent starts inside my notes via my own MCP. It queries my notes and selects up to 3 pending TODOs per project.
The Inner Loop:
- Track Creation: It scaffolds a fresh development "track" following a strict Red-Green-Refactor flow, generating spec.md, plan.md and execution metadata.
- Implementation: It implements the code by following the plan by writing failing tests first, in an isolated context and worktree .
- Verification: it review its own code against the plan, the project's style guide and ethos.
Everyone is talking about loop engineering so I thought I would share one of mine.
Every night while I sleep, my active projects (currently 8) run their own autonomous development loops. Here is how I designed "auto-work" to do verifiable coding tasks for me, allowing me to focus more on my own creative ideas:
Replying to @0xchewa
very nice, I appreciate the diagram. I've been connecting LLMs to my zettelkasten for over 2 years, one thing that I do is have a nightly agent extract out my interests and research adjacent unmentioned topics, a great way to introduce serendipity to the morning.
Revisiting an old sketch in my sketchbook inspired by Dali's Galatea of the Spheres
An image, dissolved into atoms, the picture is the illusion. Only the camera reconstitutes it, from one viewpoint.
Move, and it collapses.
Replying to @thdxr
In NYC it is a big problem for restaurants that people use bots to claim reservations and then resell them at a higher price or dont show up. Allowing this hurts the restaurants and my belly, which wants good food,
Like everyone else i'm obsessed with Angine De Poitrine
Check out their video here! youtube.com/watch?v=0Ssi-9wS…
I learned about their microtonal and polymeter ways on YT so I made a little microtonal Keytar to help me practice and get a real feel for how it all works 🎲🔺
🧵?
Replying to @Wattenberger
Media gen is already living this correction, nobody stayed happy with prompt to image for long. ComfyUI, ControlNet etc all emerged to reclaim the feedback loop.
There's still a ceiling. You can sculpt the output, but you can't think in the medium the way a painter can. Maybe that's where AI coding lands too, better than a mail slot, not quite hands in clay.
A strange thing is happening: I'm building an internet just for myself.
I now have a whole network of self-hosted sites on my VPS that I visit throughout the day, they give me reports on my interests, show results on my projects, surface connections related to my thinking, display art that has been generated, present ideas to explore, and summarize news. Many of the things I used the regular internet for are now running on my own portals, designed for me alone.
Kyle Phillips retweeted
Today, we're open sourcing the code behind "The Infinite Crate," our VST/AU plugin that lets you play with Lyria RealTime directly inside your favorite DAW! 🎧🎶
Since we released The Infinite Crate last year, it’s been used by some of our favorite artists - including a wonderful showcase with @daitomanabe in Tokyo - and being featured as a top new music tool at NAMM 2026.
We’re now fully open sourcing the plugin for developers to fork, modify, and make their own under the permissive Apache 2.0 license.
💾 Get it here: github.com/magenta/the-infin…







