From the team
Thinking out loud.
Product notes, practical guides, and honest lessons from building a proactive AI companion.
People Are Vessels of Context and Agents of Action. So Is AirJelly.
On shared context, continuity, and an intelligence that remembers our past and helps us act—while leaving our judgment and our context in our own hands.
Read article →
Archive
More from AirJelly
You know that, I am a hardware engineer
Music, games, and the everyday context beyond a chat window: a hardware engineer’s thoughts on building AI that understands us while leaving room to be ourselves.
16 min read →
Does the Company Really Own What AI Creates?
How companies can build a defensible chain of rights when people and AI create together.
24 min read →
What If Socrates Had Used AirJelly?
A thought experiment on self-knowledge, unfinished business, and how personal memory gives experience a chance to become wisdom.
8 min read →
Designing the Jellyfish: Shape, Pixels, and Light
How AirJelly draws its pixel jellyfish at runtime from bell curves, a low-resolution renderer, and a small dithered light map.
12 min read →
Exploring Proactive Agents at AirJelly: Working the Way You Do
An architectural perspective on proactive agents, user context, ongoing collaboration, and plugins that both trigger agents and provide tools.
6 min read →
The Medieval Art of Promotion for AI Founders
What Renaissance Florence can teach AI founders about winning capital, legitimacy, users, and public trust.
10 min read →
When AI Can Decide What to Do Next, What Is Software For?
Models generate possibilities; software turns possibility into order — eight principles from building a complex AI system.
13 min read →
DeepSeek Harness: A Coding Agent and a Meta Harness
DeepSeek Harness ships two things at once — an out-of-the-box Coding Agent and a pluggable Meta Harness framework. Why the second one matters more.
6 min read →
How to Extend Your Own IP Fast with Claude
Why generating pixel GIFs directly kept losing the character — and how a Pixelorama MCP plus a reusable Skill turned it into a repeatable workflow.
9 min read →
From Context Accumulation to Context Consumption
Why AirJelly is moving from sensing and storing context to helping people actively use it.
8 min read →
The Granularity Problem in GenUI: Where Should LLM Generation Stop?
Why more generation freedom made our UI worse — and why we stop the LLM at the product-semantics layer.
9 min read →
Two Flywheels: A Management Model Built for AI Startups
One big office holds down the present, one small office runs toward the future — and three roles emerge on their own.
6 min read →