OpenAI dots vs Factory for software development

September 29, 2026 - 4 minute read

OpenAI launched dots on September 29, 2026, including agents that turn customer feedback into tested pull requests. That workflow brings together finding a problem, deciding what to change, implementing it, and gathering evidence for review.

Factory offers a model-agnostic alternative for teams that want to delegate engineering work. Its Missions workflow starts from an agreed plan, while model and computer choices let teams adapt execution to their development environment.

OpenAI dots vs Factory on coding and testing

OpenAI dots. Dots can watch customer feedback, scope smaller improvements and bug fixes, build and test changes, and return pull requests with videos. The pull request and recording give the team a result to inspect before shipping.

Factory. Missions coordinate implementation and validation for bounded software projects. You agree on features, milestones, and success criteria before execution. Smaller changes can stay in ordinary Droid sessions rather than requiring the project structure of a Mission.

The Missions planning guide also describes preparing the application for verification. A user-facing project needs a scriptable way to start, accept input, or be simulated. This gives the workflow a way to check behavior rather than relying only on an implementation report.

Try each workflow on a bounded issue in your repository. Confirm that the patch addresses the original report, that checks ran against the intended revision, and that a video demonstrates the changed behavior. Record the extra instructions and corrections needed to reach an acceptable result.

Cloud computers and machine ownership

OpenAI dots. Dots have a cloud computer and browser that users can inspect. OpenAI also supports connections you approve to other devices, including a laptop.

Factory. Droid Computers provide persistent environments that retain files, installed packages, services, and configuration. Factory-managed computers pause when idle and resume when a new session targets them.

Factory's Bring Your Own Machine option lets you register a cloud VM, VPS, workstation, or on-premises server. The registered machine's existing Git credentials apply, and its maintenance and hardening remain your responsibility. This is the relevant alternative if your team wants to use an environment it already operates.

For either product, confirm the repository revision, dependencies, test data, and accessible service endpoints. Persistence can preserve useful setup and unwanted leftovers. Machine ownership also does not establish where inference runs or where connected applications send data. Those are separate parts of the deployment review.

Starting work in the background

OpenAI dots. Proactive research looks for useful work through read-only tools over connected applications when you are not actively interacting with the dot. That mode cannot send messages, modify app content, or control a browser or computer. Taking action has a separate permission boundary.

Factory. Custom Automations use configured schedules, Slack messages, GitHub events, or webhooks to start sessions. Webhook triggers remain in private preview. The run configuration specifies the task instructions, identity, execution target, and session visibility.

A Factory automation suits recurring work with a known starting point, such as a scheduled check or a new bug report. Dots' proactive research looks through connected context to discover possible work.

For a bug-report workflow, decide whether the trigger should start triage or preparation of a draft patch. Set the review owner and approval requirements before enabling it.

OpenAI dots vs Factory on model selection

OpenAI dots. Dots runs on GPT-6 Astra at launch. OpenAI's announcement does not describe a bring-your-own-model option.

Factory. Custom Models support provider keys, compatible endpoints, and local models in the CLI and desktop app. These custom configurations are not available in the hosted web or mobile products.

Factory Router offers automatic model selection that considers quality, latency, cost, and cache state. It is separate from choosing your own provider endpoint. A working custom configuration should not be assumed available through Router or on every Factory surface.

If an approved provider or local model is essential, verify that combination of model, interface, and tools against a representative task. If you prefer routed selection, evaluate Router with the same acceptance checks and record its usage separately.

Pricing and usage limits

OpenAI dots. The first dot is included in eligible Pro and Business Premium plans with an allowance for deeper work. OpenAI describes expanded limits during the first month. Tasks started or managed in Codex or ChatGPT Work retain their own usage limits.

Factory. Individual plans list Pro at $20 per month, Plus at $100, and Max at $200. Each uses independent rolling 5-hour, 7-day, and 30-day limits. Managed Droid Computers are included from Plus. Extra Usage is prepaid credit beyond the included usage.

Factory's Teams plan costs $60 per month plus $40 per seat, for up to ten seats with Pro limits per seat. Business and Enterprise use custom pricing and usage terms rather than the individual rolling-limit model.

For a software team, budget for the number of seats, the applicable allowance, and the charges after that allowance is used. Include the plan required for remote runs. Measure usage across a representative set of tasks before estimating a recurring monthly bill.

Action approvals and organizational controls

OpenAI dots. Custom Rules can permit an action, require approval, or block it. Activity View exposes progress, and action auto-review checks proposed actions against instructions and safety requirements. Certain sensitive actions, including password changes, always remain with the user.

Factory. Enterprise Controls distinguish enforced organizational boundaries from defaults that users can choose locally. Hard controls remain authoritative over more local settings. A preferred model and an enforced model restriction are different controls.

Test permissions against a specific workflow. Name the repository, allowed tools, execution environment, and review owner. Confirm that the configuration blocks an action the agent should not take and asks for approval at the intended point.

A team with an approved model provider and an existing development machine has concrete reasons to try Factory. A team delegating research, proposals, and engineering work together may also value dots' broader scope. In either case, keep merge and deployment approval with the people responsible for the service.