GitHub Copilot Dev Days — Boston
AICamp · Microsoft New England (NERD) · 2026-04-29
Table of Contents
Overview
Community-led developer event on AI-assisted coding with GitHub Copilot. Three-hour structured program (talks + hands-on build), followed by science-fair-style demos in The Garage Makerspace. Hosted by AICamp in the Microsoft NERD facility.
Agenda
| Time (EDT) | Block |
|---|---|
| 15:00–15:15 | Welcome, intro |
| 15:15–16:00 | GitHub Copilot, CLI tools, developer tooling |
| 16:00–17:00 | Hands-on build session |
| 17:00 | Networking, food (continues into evening) |
| 17:15–19:00 | Community Science Fair, demos in The Garage |
Speakers
- Chris Templeman — Microsoft
- Kayla Cinnamon — Microsoft. DX surface for terminal/CLI tooling.
- Maddy Montaquila — Microsoft, .NET MAUI PM lineage; mobile + cross-platform dev tooling.
- Kranthi — affiliation pending in source.
Lineup was marked "updating" on the AICamp page; treat any retrospective attribution against actual delivered talks, not the pre-event list.
Topics
Copilot as harness, not feature
Reading Copilot through the agent-harness catalog (~40 named systems, including Claude Code, Aider, Gemini CLI, Cursor, guile-sage). Copilot occupies the IDE-resident + CLI quadrant; the event's "CLI tools" segment is the relevant probe point for terminal-resident agent surface.
Tool surface vs editor surface
Copilot's editor integration is well-mapped; the CLI surface is the newer contract. Where does it land on:
- Provenance: how is the model invocation logged, reproducible, attributable?
- Refinement loop: is the CLI an autonomous loop or a single-shot completion?
- Blast radius: filesystem writes? shell exec? Tool gating?
These are the same questions applied to other harnesses; consistency of questioning is the contract.
Hands-on build session
Format-dependent. If the build is a guided "build X with Copilot," the elicited contract surface is shallow. If participants choose the target, the session becomes a small natural experiment on what people reach for when given a production-grade assistant.
Community Science Fair
Likely the higher-signal portion. Non-vendor demos reveal where Copilot adoption has landed in practice, including failure modes that don't surface in vendor talks.
Tie-ins to current research
- Agent harness ecosystem audit (catalogue, five-column feature matrix). Event provides a vendor-side data point against the more research-y reference systems (guile-sage, emacs-mcp-maximalist).
- aq gossip protocol assumptions about ambient agent presence intersect with whatever Copilot's CLI does about persistent state.
- Governance tuple framing [persona:agent:reviewer@env(project:workspace)]: test whether Copilot's session model maps cleanly or breaks the tuple.
Refutation conditions
The agent-harness-catalog framing for this event is wrong if:
- The CLI segment is purely a feature walkthrough with no exposed contract surface (no governance, no refusal modes, no provenance).
- The build session has no degrees of freedom for participant target choice.
- The science fair is dominated by Microsoft-internal demos rather than community work.
In those cases the event reduces to a networking visit with no research deliverable, and the org entry stays a logistics record.
Notes from attending
: - : - : - : - Science Fair standouts - Conversations worth following up
Logistics
- AICamp Discord (events + community): https://discord.gg/2DySXfHTFX
- Submit speaker topics for future AICamp events: https://forms.gle/JkMt91CZRtoJBSFUA
Sources
- AICamp event page: https://www.aicamp.ai/event/eventdetails/W2026042913
- Page fetched and diffed against initial pull on 2026-05-11; no drift in agenda, venue, speaker list, or RSVP count (77).