Makerkit.dev's answer

Indie Hackers and Companies who want to launch quickly, without compromising on quality.

Harbor ML's answer:

Harbor serves companies building physical AI systems, including:

Robotics companies (industrial, logistics, manufacturing)

Autonomous vehicle developers

Consumer AI hardware manufacturers

Wearable AI platforms

Enterprise computer vision systems

These are typically:

AI-first startups building embodied systems

Mid-to-large enterprises integrating robotics

Frontier AI companies expanding into physical environments This is a technical, infrastructure-focused audience — not casual developers.

Makerkit.dev's answer

Makerkit uses Next.js 15 (App Router), Supabase, React.js, Typescript and Stripe.

Harbor ML's answer:

At a high level, Harbor ML is built on five core technology layers:

1. High-throughput Data Ingestion

Real-time sensor and video ingestion

Scalable distributed storage

API-based data pipelines

1. Video Infrastructure Stack

Media distribution systems

Edge ingestion systems

Hardware integration pipelines

1. AI Pre-Labeling Models

Computer vision models

Object detection systems

Edge case detection models

Foundation model integration

1. RLHF Infrastructure

Human-in-the-loop annotation systems

Quality control tooling

Contributor ranking systems

Feedback reinforcement pipelines

1. API Delivery Layer

Dataset versioning

Enterprise API access

Secure dataset distribution

Monitoring & model feedback loops

The technical backbone likely includes:

Distributed systems architecture

Cloud-native infrastructure

Machine learning pipelines

Video processing frameworks

Secure API gateways

Makerkit.dev's answer

Makerkit stands out by offering a truly modular architecture built with Turborepo, where core features like auth, billing, and notifications live in their own packages for better maintainability.

While most starters lock you into specific patterns or providers, Makerkit gives you flexibility with a multi-account system supporting both B2B and B2C scenarios, provider-agnostic billing, and edge-ready deployment options.

Beyond the basics, it includes production-ready features like multi-factor auth, real-time notifications, and team permissions - all built with Supabase, TypeScript, React Query, and modern tooling to make development a genuine pleasure.

Harbor ML's answer:

Harbor ML is not an annotation company.

It is the infrastructure layer for RLHF in physical AI.

Most players in robotics data operate at one layer:

Data labeling

Tooling

AI models

Workforce marketplaces

Harbor ML controls the entire pipeline:

Capture → Distribution → Recruitment → RLHF → Delivery

That vertical integration is rare.

The second differentiator is its media infrastructure advantage. Harbor doesn’t just wait for customers to upload data — it operates a vertically integrated media and distribution stack to source both data and contributors at scale.

Third, Harbor is specifically built for physical AI, not text or generic vision models. Physical AI requires:

High-fidelity sensor ingestion

Real-world edge cases

Human interpretation of spatial and behavioral context

Harbor industrializes this through a proprietary RLHF pipeline.

In short: Harbor is building the AWS-equivalent infrastructure layer for robotics data — not a service business.

Makerkit.dev's answer

While other starters give you basic auth and a dashboard, Makerkit provides a genuinely modular foundation with the real features SaaS products need - like multi-factor auth, team permissions, real-time notifications, and provider-agnostic billing, all organized in clean, maintainable packages using Turborepo.

You get a first-class developer experience with TypeScript, React Query, and modern tooling, plus the flexibility to support both B2B and B2C scenarios, different payment providers, and edge deployment options.

Best of all, Makerkit is actively maintained with regular updates and responsive support, so you're building on a foundation that grows with your needs rather than painting yourself into a corner.

Harbor ML's answer:

Because Harbor solves the real bottleneck: scalable, high-fidelity real-world data with human feedback baked in.

Compared to traditional annotation firms:

Harbor offers full infrastructure, not just labor.

Harbor combines AI pre-labeling + human refinement.

Harbor builds recurring, API-delivered datasets.

Compared to pure AI model companies:

Harbor doesn’t compete on the model.

It enables every model company to perform better in reality.

Compared to marketplaces:

Harbor focuses on quality control, vetting, and RLHF logic — not just gig labor.

The core advantage for customers:

Faster deployment

Higher real-world reliability

Lower long-term data costs

Continuous dataset improvement

If you’re building physical AI and care about deployment performance, Harbor reduces failure risk.

And in robotics, deployment failure is expensive.

Harbor ML's answer:

The story starts with a simple realization:

Robots fail not because models are weak — but because they lack grounded, real-world training data.

Simulation works up to a point. But the real world is messy. Sensor noise. Lighting shifts. Human unpredictability. Edge cases everywhere.

The founders recognized that physical AI would follow the same path as language models:

First breakthrough models. Then realization that data quality and RLHF determine performance. Then a massive need for infrastructure.

OpenAI had RLHF for text.

Physical AI had nothing comparable.

Harbor ML was created to industrialize RLHF for embodied intelligence.

Instead of treating data as a service, Harbor treats it as infrastructure — building the essential supply chain for physical intelligence.

The long-term ambition:

Become the default data layer powering every robot and embodied AI system globally.

Harbor ML's answer:

Harbor is a strategic solution partner to:

Adobe

IBM

Beyond that, the target customer profile would include:

Robotics manufacturers

Autonomous vehicle platforms

Wearable AI companies

Industrial automation firms

Enterprise AI system integrators

At pre-seed stage, it’s important to be precise:

If Harbor has signed enterprise partners, name them clearly. If not, position them as active pipeline targets rather than implied customers.

Tier-1 investors will probe this immediately.

Clarity builds trust.