What we're NOT building (and why that's a feature)

Product strategy is usually framed in terms of what you ARE building. The roadmap. The features shipping next.

The less-discussed half: what you're DELIBERATELY not building. That's where the focus comes from.

Here's what Avery NXR is not — and why each "no" makes the product better.

We're not building a chatbot

Avery NXR doesn't have a chat interface. You don't open the app and type "do this for me."

Why not: because the use case we're optimizing for is recurring operational work, not one-off conversations. Adding a chatbot would dilute the design center. Users would expect chatbot behavior. We'd build for two different audiences poorly instead of one audience well.

If you want a chatbot, use Claude, ChatGPT, or Perplexity. They're great at chatbot. We're built for the work that happens AFTER you stop chatting.

We're not building a cloud LLM

We use cloud LLMs (via Consult Mode + BYOK keys) but we're not building our own. We don't have a "Avery LLM" or fine-tuned in-house model that we push as the default.

Why not: because frontier LLM research is a different business than agent platforms. Anthropic, OpenAI, Google, Meta, Mistral are spending billions on the model layer. The right call for us is to use the best of what they ship.

Local model layer: Ollama + open-source models. Cloud escalation: Claude / GPT / Gemini via your keys. Not "our" model. Better models for the work.

We're not building a coding agent

Avery NXR has shell access as one of its 59 capabilities, and agents can run code. But we're not positioning as a coding agent platform.

Why not: because Cursor, Cognition (Devin), Replit Agent, Magic.dev, OpenHands, etc. are building coding-specific agents at depth we won't match. They have audiences and product motions we don't.

Avery NXR is for operational agents — the work surrounding code, not code itself. Meeting notes, support triage, pipeline digests, resume screening. The boring operations.

If you want a coding agent, use one of the products above. They're great. We're built for adjacent work.

We're not building voice agents

No voice interface in Avery NXR. No phone calls. No voice-driven configuration.

Why not: because voice agent platforms (Sierra, Decagon, Bland AI, Vapi) are building deep voice expertise — telephony, ASR, TTS, latency optimization, conversation design. We won't beat them on voice.

Avery NXR is for background agents. The work happens. You read the output. No voice required.

We're not building a multi-tenant cloud SaaS as the primary product

Avery NXR Free Desktop runs on your laptop. Avery NXR Pro deploys to YOUR infrastructure (Vercel, Railway, SSH). Avery NXR Enterprise is on-prem capable.

What we're NOT doing: building "Avery Cloud" where everyone's agents run on our centralized infrastructure with multi-tenant data.

Why not: because the privacy + cost + control properties of local-first are core to why Avery NXR exists. If we built a multi-tenant cloud as the default, we'd be Lindy or Relevance AI. The world doesn't need another cloud-first agent platform. It needs a local-first one.

We're not building marketplaces for community agents

There's no Avery Agent Marketplace where users buy and sell agents.

Why not: because agent quality depends heavily on configuration matching specific business context. A marketplaced agent built for a 5-person startup's HubSpot is unlikely to work for a 500-person enterprise's Salesforce. Marketplaces create the illusion of plug-and-play that doesn't survive contact with real business contexts.

What we DO ship: 7 production templates that we maintain. We control the quality. Users fork the templates for their specific context.

We're not building consumer-facing AI

No "Avery for personal life" version. No relationship advice agent. No meditation agent.

Why not: because consumer AI is a different market motion (D2C, ads, attention economics) from B2B operational AI (long sales cycles, security review, recurring revenue). Trying to do both well is harder than doing one well.

We chose B2B operational. Consumer AI is a different fight for a different company.

We're not building model fine-tuning workflows

Avery NXR doesn't ship a UI to fine-tune models on your data.

Why not: because fine-tuning is a complex workflow that needs ML expertise to do well. Doing it badly produces worse output than not fine-tuning at all. We don't want to be the platform that gave non-ML teams a fine-tuning button that hurt their results.

What we do instead: support a wide variety of base models so users can pick the right tool for the job. Plus Consult Mode for escalation to frontier when local models aren't sufficient.

Why "not building" is a feature

Every "no" above is a deliberate scope choice. Each one increases focus on the thing we ARE building: a local-first, operational, background, transparent AI agent platform.

When competitors try to be everything — chatbot + agent + coding + voice + marketplace + consumer + B2B + cloud + local — they end up being mediocre at all of it.

When a product is clear about what it ISN'T, the audience that fits the product self-selects in. The audience that doesn't fit goes elsewhere. Everyone is better served.

What this means for users

If you're evaluating Avery NXR and notice we don't do X, check whether X is on this "deliberately not building" list:

→ Chatbot conversation — no → Our own LLM — no → Coding agents — no → Voice agents — no → Multi-tenant cloud SaaS — no → Agent marketplace — no → Consumer AI — no → Fine-tuning UI — no

If your use case requires any of these, we're not the right tool. Pick something else, no hard feelings.

If your use case is operational, recurring, background, agent-driven work where data residency and cost predictability matter — we're built for you specifically.

The principle

"You can't build everything" is true. "You shouldn't build everything" is also true and underdiscussed.

The companies that survive long-term are the ones that make peace with their scope, get focused, and beat broader competitors by being deeper in the slice they chose.

Avery NXR is our slice. We're committing to it.

→ avery.software — Free Desktop tier. The agent platform that knows what it's NOT.