Builds & prototypes
DataDive AI
An on-demand AI data scientist on Twitter.
Receipts
An AI interviewer that turns a project conversation into interview flashcards, with supporting quotes kept separate from the written story.
HomeLore
A voice-first family-memory prototype that keeps original words, generated narration and unknowns distinct.
Paper IRL
Turns saved X bookmarks and linked articles into a finite, private edition that can be printed.
AI video editing system
An AI video editor for my podcast and cricket footage.
Onchain Sentinel
Verifiable onchain leaderboards for engagement on X, with autonomous creator rewards on Base.
Onchain raffle agent
An agent for running verifiable, autonomous global onchain raffles in under 60 seconds.
Universal creator monetization agent
Lets brands work directly with creators and pay per view, regardless of creator size.
ChatRadar (Telegram Bot)
A Telegram bot that summarizes unread messages, tracks topics across groups, and sends on-demand DMs.
AI Manager / Maya
A frontend product-strategy review prototype with predefined scenarios, evaluation design and failure states. No achieved evaluation results are claimed.
Chop
A product in development where creators remix a brand’s existing assets into short-form content.
Khi-Khi: Emotion sharing beyond close proximity
A wearable concept that detects laughter through heartbeat, sound and motion sensors and triggers a remote module to tickle a friend.
Heimdall
A live personal prototype for reading a finite selection of X posts around a current goal.
Local AI Health Companion
An experimental on-device health-information companion with chat, a symptom journal and a PDF handoff.
Work Evidence
An independent concept prototype that pairs a role-relevant work sample with evidence a candidate reviews and approves for employers.
Super Over
A mobile cricket batting-game prototype: one thumb, six balls and a target to chase.
Workflow Agency
A scaffold for agents to research competitors, verify findings and prepare them for publication, with delivery checks visible.
Local Statement Analyzer
A local bank-statement analysis prototype that uses on-device models for categorization and deterministic tools for arithmetic.
X Radio
A personal timeline-to-podcast prototype that turns X posts into speakable text, audio episodes and an RSS feed.
Veil
A live browser-only CSV prototype for reviewing bank transactions and exporting a smaller selection of fields.
Pinboard for Claude
A browser extension prototype for pinning snippets from Claude conversations onto a sticky-note board.
One Over - real cricket
I built a browser game from my real nets footage. Choose where to bowl and watch me play the shot.
Product thoughts
Making agent work steerable before the change lands
A proposal for adding an instruction to a specific step of an agent task and seeing whether the agent can still include it.
Making agent work steerable before the change lands
A proposal for adding an instruction to a specific step of an agent task and seeing whether the agent can still include it.
The change was small
I asked an agent to put a holding page on my portfolio. While it was working, I wanted to add LinkedIn alongside my X profile. The change was small. Knowing whether the instruction would reach the right part of the work was harder.
I could send another message, but I could not easily see whether the agent was still preparing the page, had already updated the links, or was about to deploy. I wanted to steer the work without stopping it and without triggering a second deployment.
Attach the instruction to the work
I would start with a short, visible map of the current turn. A new instruction could refer to one of its steps. I could ask to include LinkedIn in the profile-links step and see an acknowledgment beside the affected work.
When I send an instruction, I want to know whether the agent can still include it in the deployed result. An acknowledgment should tell me that, beyond confirming receipt.
The first MVP could be text: the steps, the current position and the acknowledgment beside the affected step. A visual component can follow once the underlying state is reliable.
Proposed interaction
A flow to test. In this example, a new instruction targets the profile-links step.
- Read current state
- Prepare page
- Update profile links
- Review
- Deploy
- Received
- The system has captured the instruction.
- Included
- The instruction is attached to work that can still incorporate it.
- Needs another pass
- The relevant step is already complete, so this change needs another pass.
Where a change needs another pass
I would define when a change needs another pass. Updating a local file and deploying a public page are different events. The interface should show whether there is still time to include the change, without making me understand the agent’s internal machinery.
I would keep this as a proposal until the steps and acknowledgments reliably match the work. A reassuring but inaccurate status would leave me with the same uncertainty.
What I would test
I would give people a small change while an agent task is in flight. I would check whether they can tell if it was included, whether they repeat themselves less, and how often the task needs another deployment. I would also check whether maintaining the map slows the work down.
I would also check whether the map and acknowledgments stay accurate when an instruction arrives after its step has finished.
The unit of value is finishing
Paper IRL explores whether turning saved X bookmarks into finite, private editions helps readers return to material they already chose.
The unit of value is finishing
Paper IRL explores whether turning saved X bookmarks into finite, private editions helps readers return to material they already chose.
An edition that can be finished
People save posts because they want to return to them. The collection grows, but reading competes with the same feeds and interruptions that created the collection. Paper IRL starts with an edition that can be finished.
The reader has already chosen the material by saving it. I want to turn it into a private physical edition they can sit down and finish.
What that changes
The saved material carries the reader’s intent, so the early product does not need to invent a ranking score. Source links and context help preserve what was saved. The edition needs to be readable and finite.
I would judge the early experience by whether a reader can sit down with an edition and finish it. I would add more material or more ways to browse only if they help with that. I still need to test whether this changes reading behavior.
Printing and delivery would remove another chore. I can promise them only after the order path works; producing a print-ready file is one step.
Prototype evidence and the next test
The prototype records describe HTML and PDF files with access scoped to the owner, an explicit sample mode and output that preserves the source material. Whether readers finish more material, and whether paid fulfillment works, remain untested.
A useful next experiment would compare what people finish from a small edition with what they revisit in their ordinary save queue. I would also ask whether the edition earns a recurring place in their week.
Ideas & demos
Why I started Creatorbase
Why I started Creatorbase, who it serves and the creator monetisation problem behind it.
- Product strategy
- Creator economy
Why I started Creatorbase
Why I started Creatorbase, who it serves and the creator monetisation problem behind it.
Chop: content distribution and creator pay
A clipping workflow where brands amplify existing content and creators earn per view. The demo introduces Chop alongside original UGC campaigns.
- Creator incentives
- Distribution
Chop: content distribution and creator pay
A clipping workflow where brands amplify existing content and creators earn per view. The demo introduces Chop alongside original UGC campaigns.
Pull it up Jamie
A desk device concept that researches a spoken claim in the background, then signals when the results are ready so the conversation can continue.
A product idea illustrated with a short clip.
- Voice interfaces
- Product ideas
Pull it up Jamie
A desk device concept that researches a spoken claim in the background, then signals when the results are ready so the conversation can continue.
A product idea illustrated with a short clip.
Onchain raffle agent
An agent demo for creating and running onchain raffles, built with AgentKit, Replit and Alchemy.
- Agent workflows
- Onchain products
Onchain raffle agent
An agent demo for creating and running onchain raffles, built with AgentKit, Replit and Alchemy.
An agent for creator rewards
An early demo of an onchain agent for rewarding content creation, shared from ETHIndia.
- Creator economy
- Agent workflows
An agent for creator rewards
An early demo of an onchain agent for rewarding content creation, shared from ETHIndia.