2025 - 2026
Timeline
1 Designer
8 Developers
Team
Led end-to-end design of the Augment IDE product from March 2025 to January 2026 as the sole designer supporting a team of 8 engineers. Shipped at the pace of a fast-moving market, maintaining feature parity with competitors while the market redefined itself month to month.
Sole Designer
Augment Code AI dev platform built for large, complex codebases, powered by a Context Engine that indexes up to 500K files across dozens of repositories so its agents can reason about changes spanning an entire system rather than a single file. Developers work with it directly in VS Code and JetBrains — chat and an in-editor coding agent.
Agentic IDE Extension
Impact
Augment brings AI to the whole engineering organization — not just the individual developer — powered by a context engine that reasons across entire codebases
22M
revenue (arr)
60K
users
2
IDEs
Product Evolution
Shipping at Market Speed
Context & Model Selection
Context references have multiplied over the past year and a half of agentic interaction, and the prompt input is where that accumulation lands. Adding configuration controls, setting sensible defaults, and deciding what to nest has been a continuously moving target.
​
-
Adding context references
-
MCP connectivity
-
Model price evolution
-
Model thinking selection
Thread Consolidation
As trust in agents grew, fewer users wanted to inspect the reasoning behind every decision. But that shift wasn't uniform — it left two audiences to serve at once: those who still reviewed each step manually, and those who were ready to let agents run.​
​
-
Code snippet exposure
-
Tool call display evolution
-
Turn summaries
Prompting
Agents take real time to work. Once users trusted the output, they stopped waiting for each turn to finish — queuing up a series of prompts instead and checking back when it suited their own workflow.​
​
-
Prompt enhancement
-
Prompt editing inline
-
Queued prompts
-
Agent task and spec creation
Agent Monitoring
Once the infrastructure supported it, running multiple agents in parallel became possible — and that raised a new requirement: knowing where each one stood. Notifications, alerts, and a way to see every agent's status at a glance suddenly mattered.
​
-
Agentic chat
-
Auto vs manual tool approvals
-
Multi-agent monitoring
Code Review
Code review has been the most contested checkpoint in the SDLC since agentic coding arrived. The move from approving changes line by line to handing agents full control is the single biggest shift in how humans and agents work together.
​
-
In thread diff review
-
Next edit
-
Diff list
-
Auto approve
Demo
Product Walkthrough
