REVIEWSCOPE — STAGE 1

Competitive Analysis — OpenAI–Ona Threat Refresh

Product: gblock-party — always-on, BYO-client, mobile-first control plane for AI coding agents · Category: Developer Tools SaaS — AI Agent Orchestration · Date: 2026-06-13

Trigger: OpenAI acquires Ona (formerly Gitpod), announced 2026-06-11. This page proposes research scope only — no findings yet. Approve the scope to unlock Stage 2 research.

Why this refresh — what the acquisition changes

The existing canonical report (research/competitive-analysis.md, 2026-05-25) named first-party encroachment as the #1 competitive risk — but judged it survivable because every first-party mobile/agent play was single-agent, single-provider. The OpenAI–Ona deal is the first data point that breaks that comfort:

Ona's persistent VPC-isolated environments + governance (audit trails, scoped credentials, kernel-level policy) now back Codex (~5M weekly active users).
It directly targets gblock-party's Gap 2 — session persistence without cloud lock-in ("agents keep working after you close the laptop").
Analysts read it as defensive vs. Anthropic, mirroring Anthropic's May 2026 self-hosted sandboxes for Claude Managed Agents. The labs are now consolidating infrastructure + governance, not just models.

Net: the existing Ona entry ("ex-Gitpod, enterprise-only, $41M, own-agent ecosystem") is now stale and must be re-framed as an OpenAI-owned Codex infrastructure layer.

Approved scope

Breadth — Focused (your selection): treat the acquisition as the trigger and re-assess the first-party platform layer only — OpenAI/Codex + Ona, Anthropic/Claude Managed Agents, Cursor, GitHub Copilot. The 27-competitor indie/OSS landscape from 2026-05-25 is not re-derived; those entries remain current.

Frameworks — all four (your selection): SWOT · Porter's Five Forces · Feature/Pricing Matrix · Strategic Group Map.

Frameworks queued

SWOT selected

gblock-party S/W/O/T specifically against OpenAI–Ona and the consolidating first-party layer. Directly answers "is this a threat, and where is it survivable?"

Porter's Five Forces selected

How a $40B-funded entrant absorbing the infra layer shifts industry structure — entrant power, substitutes, buyer power, supplier (model) power, rivalry.

Feature / Pricing Matrix selected

Update the existing matrix row-set with Codex+Ona (persistence, VPC isolation, governance, pricing) vs gblock-party and the field.

Strategic Group Map selected

Re-cluster post-acquisition to show where whitespace survives: solo-dev / own-host / model-agnostic / human-in-the-loop, after the labs converge on enterprise VPC autonomy.

What the synthesis will change in the canonical report

Stage 3 will update, not overwrite, research/competitive-analysis.md. Anticipated edits (to be evidenced in Stage 2):

Section	Currently says (2026-05-25)	Refresh direction
Ona entry (Platform Incumbents)	"ex-Gitpod, enterprise-only, ~72 team, $41M, OCU pricing, own agent ecosystem only."	OpenAI-owned (June 2026); becomes Codex's persistence + governance infra layer; re-evaluate model-agnosticism and whether a self-serve tier appears.
Summary — "biggest risk"	First-party encroachment exists but "all remain single-agent, single-provider."	First-party players now acquiring infra + governance; persistence gap (Gap 2) is being closed by a vendor. Re-rank risk severity.
Gap 2 — Session persistence	"No managed solution offers device-independent persistence with BYO-client flexibility."	Codex+Ona and Claude Managed Agents now offer persistence — but vendor-locked. Re-state gap as persistence + BYO-client + model-agnostic, narrower and sharper.
Evidence matrix + Assumption register	"First-party mobile solutions are single-agent only" (High).	Downgrade/qualify; add acquisition + Claude Managed Agents evidence rows.

Source plan & coverage gaps

Available sources: existing canonical research (competitive-analysis.md, icp.md, positioning.md, journey-map.md BYO-client landscape) + web search (last-12-months priority): acquisition coverage (InfoWorld, Bloomberg, TNW, Techzine, MLQ, Built In), ona.com, Anthropic Claude Managed Agents / self-hosted sandbox docs, Codex WAU + pricing, current Cursor/Copilot state.

Open questions to resolve in Stage 2

Codex+Ona combined pricing and whether a self-serve / individual tier survives the deal.
Does Ona stay model-agnostic, or become Codex-locked (which would kill it as neutral infra)?
Claude Managed Agents current GA status, self-host details, pricing.
Ship timeline: Codex-on-Ona GA vs. enterprise-preview only.

Assumptions & confidence

Assumption	Confidence	Status
Acquisition announced 2026-06-11; terms undisclosed	High	Multiple outlets
Codex ~5M weekly active users	Medium	Press single-source; verify
Ona folded into Codex (not kept fully neutral)	Low	Inference — open question
Existing OSS/indie entries still current	Medium	Assumed; out of focused scope

Output paths

Working packet (Stage 2): research/_working/preliminary-competitive-analysis-research.md
Framework intermediates (Stage 2): research/competitive-analysis-swot.md, …-porter-five-forces.md, …-feature-pricing-matrix.md, …-strategic-group-map.md
Canonical (Stage 3): update research/competitive-analysis.md + research/competitive-analysis-search-log.md
This alignment page: alignment/competitive-analysis-openai-ona.html

Approval gate — research scope

Approve to unlock Stage 2 research

No synthesized research, ranking, or canonical writes happen until this scope is approved. Confirm or adjust:

1. Breadth. Focused first-party-layer refresh (vs. full 27-competitor re-run). → confirmed in intake; change here if you want the full re-run.

2. Frameworks. All four (SWOT, Porter, Feature/Pricing, Strategic Group Map). → confirmed in intake.

3. Canonical handling. Update the existing report in place (revise Ona entry, risk thesis, Gap 2, matrices) rather than create a new file. → default; say if you'd prefer a separate dated report.

4. Comparators. Include Anthropic Claude Managed Agents + Cursor + Copilot alongside Codex+Ona. → default; add/remove any.

Stage 1 of 3 (Scope → Research → Finalize). This page renders proposed scope only and contains no synthesized findings. On approval, framework steps + a synthesis step are queued to tasks/todo.md and driven by /exec.