Elena Shuvalova

Growth Product (antiques & collectibles)

[Elena Shuvalova @creshka]

Thrilled to share that I’ve been accepted into Delta by Live The Residency — a program that brings together founders shaping the future of art, culture, and technology.
As I’m building Art AI Advisor — the first personal AI art consultant.
Applications are still open

[Elena Shuvalova @creshka]

Full technical breakdown is on Substack. https://t.co/BA2cJWxNUY

[Elena Shuvalova @creshka]

Building self-optimizing workflows in Codex using fuzzy logic and genetic algorithms @OpenAI

[Elena Shuvalova @creshka]

A workflow is only as good as its weakest required criterion.

Strong domain fit should not compensate for weak source trust or poor risk reduction.

[Elena Shuvalova @creshka]

@thsottiaux One thing that does not feel delightful yet is how hard it is to explicitly choose the current file as context.

When I work in Codex, I often want to ask something very specific about the file I am currently looking at.

[Elena Shuvalova @creshka]

https://t.co/lz9Gk0QvjT

[Elena Shuvalova @creshka]

Most AI agent setups eventually run into the same problem: the more capable the system becomes, the more context it wants to load. https://t.co/lz9Gk0QvjT

[Elena Shuvalova @creshka]

The fix I built was deliberately boring: a lightweight fuzzy routing process for choosing skills before doing the work.

[Elena Shuvalova @creshka]

What My AI Agents Actually Cost and the Unit Economics That Say If They’re Worth It https://t.co/yWgCFmmolJ

[Elena Shuvalova @creshka]

I have serious FOMO right now.
There’s a pool at the dacha. The cats are there. The tomatoes are blooming.
And I’m in the city configuring Hermes @NousResearch with my fuzzy logic instructions.

[Elena Shuvalova @creshka]

Read how I made it on Substack: https://t.co/wABNsbVLEh

[Elena Shuvalova @creshka]

I already had IBM Selectric sounds in Obsidian. Then I wanted the same satisfying click in Codex.

So I built a tiny Mac app that makes my AI workspace clack like an old typewriter.

[Elena Shuvalova @creshka]

How Anthropic Managed Agent Helped Me Build a Self-Improving DevOps System by Ilia Popov https://t.co/zd3gchEWUB

[Max Postnikov @postnikov]

What comes next after prompts and skills?
I have the answer.
Blueprints. Transferable AI Workflows.

https://t.co/qfr9z8W0aR

@karpathy @AnthropicAI @simonw