Hieu Phung (Henry)
Hieu Phung (Henry) is an individual AI/ML engineer, writer, and open-source developer rather than a company. His public work spans machine-learning education and experimentation, personal writing about running and coffee, and Dely, an open-source workflow for taking software changes through design approval, implementation, independent review, and pull-request preparation.
At a Glance
- Individual software developers
- Engineering teams using multiple AI coding assistants
- Open-source projects and repositories needing controlled AI-assisted delivery
- Developers seeking tool-agnostic implementation and review workflows
AI Tools by Hieu Phung (Henry)
(1)
Dely
AI Agent Workflow Plugin
Discussions
No discussions yet
Be the first to start a discussion about Hieu Phung (Henry)
Latest News
Dely 0.24.1 released with a setup documentation fix explaining personal pin lookup in linked worktrees.
Dely 0.24.0 moved team and personal harness pins out of AGENTS.md into .dely configuration, added strict pin validation, and introduced dely pins.
Dely 0.23.0 added Pi support, a harness-derived role table, and Copilot, Antigravity, and Grok support.
Dely 0.22.0 added OMP support with a spec-carried model pin.
Products & Services
Open-source, MIT-licensed, Orca-supervised delivery protocol/plugin for coding agents. It turns a request into an approved design contract, starts isolated implementer and reviewer sessions, supports sequential implementation and independent review, and prepares a pull request for the user to merge.
Personal publishing site containing Hieu's writing on machine learning, running, coffee brewing, and personal stories, including articles on image-aesthetics assessment with AI and marathon training.
Market Position
Dely positions itself as an open-source, cross-harness control and delivery protocol rather than a standalone coding model or IDE. Its differentiation is the explicit design-contract gate, isolated implementer/reviewer sessions, Orca supervision, and support for multiple coding-agent harnesses; this contrasts with single-agent coding skills and workflows such as Superpowers that focus more narrowly on one agent workflow.
Founding Story
Hieu said he developed Dely from his own multi-tool coding workflow: instead of carrying conversation context between coding agents, a planning session produces a short design contract containing the goal, acceptance criteria, scope, and exclusions; an implementer and an independent reviewer receive the contract and repository/diff separately. He then automated that handoff with a plugin requiring Orca.
Business Model
Revenue Model
Dely is distributed as public MIT-licensed open-source software through GitHub. The reviewed materials describe installation from Git repositories/plugin marketplaces and do not describe a paid subscription, API metering, or commercial license.
Target Markets
- Individual software developers
- Engineering teams using multiple AI coding assistants
- Open-source projects and repositories needing controlled AI-assisted delivery
- Developers seeking tool-agnostic implementation and review workflows
- Multi-agent software development across different coding-agent harnesses
- Feature and bug delivery where scope, acceptance criteria, and review independence matter
- Teams or individual developers using Claude Code, Codex CLI, Cursor Agent CLI, GitHub Copilot CLI, Antigravity CLI, Grok Build, OMP, or Pi
- AI-assisted implementation with an explicit human approval gate and a separate review pass