Blog

Page 4 of the Notis archive: practical systems for delegating work, capturing ideas, and running a company with an AI intern.

Build in Public Case Studies Blueprints Tutorials Analysis Comparison

Reviewed by Florian ParisetComparisonSep 22, 2026

7 Reclaim AI Alternatives, Compared by What They Actually Charge You For

Reclaim bills per seat, then bills again for meeting attendees who never signed up, and rents you a scheduling horizon by tier. Seven alternatives compared on the unit that decides your invoice, with every price read from the vendor's own pricing page.

Reviewed by Florian ParisetComparisonSep 22, 2026

Tana Alternatives: 7 Options Ranked by What They Actually Bill You For

Tana's Pro price is an early-bird rate against a $30 regular price, and AI is billed in credits on top of the seat. Seven alternatives compared on the axis that decides your bill, with every price taken from the vendor's own page.

Reviewed by Florian ParisetComparisonSep 21, 2026

7 superwhisper Alternatives That Do Not Charge You for Your Own API Key (2026)

superwhisper's free tier is capped by model class, not minutes, and bringing your own API key is a Pro feature. Seven alternatives compared on what the price is actually charged against, with every price read from the vendor's own site.

Reviewed by Florian ParisetComparisonSep 20, 2026

7 Notion AI Alternatives, Ranked by What They Actually Charge You For

Notion AI only ships on Business and above at $20 per member per month, plus $10 per 1,000 agent credits. Seven alternatives compared on the unit each one bills against, with every price taken from the vendor's own page.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

Docs Your Agent Can Read: Why the Notis Help Centre Runs on Mintlify

Docs-as-code means a coding agent can update the Help Centre in the same commit as the feature — and clean structured output means models can read it accurately.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

Boring Hosting Is a Feature: How Notis Runs on Render

What actually runs on Render, the thirty-line relay that solved our static-IP problem, and why environment groups beat per-service configuration every time.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

How Notis Delivers an AI Assistant Inside WhatsApp With Twilio

WhatsApp is not a chat widget. Templates, session windows, media and delivery callbacks — and why failed delivery is product logic, not telemetry.

Reviewed by Florian ParisetComparisonSep 19, 2026

Best AI Chat Apps for Founders in 2026: Conversation Isn’t Delegation

The best AI chat app isn’t the one with the smartest answer. For founders, the real test is whether it can move work across email, calendar, and CRM.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

Turning Web Pages Into Agent Context With Firecrawl

Raw HTML is not context. How Notis uses Firecrawl to turn pages into Markdown an agent can think with — and the un-metered tool bug that cost us real money.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

Social Proof You Can Click: How We Run Our Testimonial Wall With Senja

Every quote on the Notis homepage links to a public source. Here is how we collect them with Senja, and why we self-host a pinned copy of the widget loader.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

The Emails an AI Assistant Owes You: How Notis Used Customer.io

Agents fail quietly. Here is the full inventory of operational emails Notis sent through Customer.io, and why triggers belong in code while copy belongs in the ESP.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

One State Layer for an AI Assistant: How Notis Uses Supabase

Conversations, documents, automations and durable agent sessions all live in one Supabase project — and why the file-sharing trade-off is worth defending out loud.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

You Can't Debug an Agent From a Log Line: How We Traced Notis With Langfuse

How Notis used Langfuse to see what an agent actually did, why we wrapped the SDK to enforce masking, and the thread leak that taught us to cache clients. Notis is now replacing it with run records in its own database.

Reviewed by Florian ParisetBuild in PublicSep 19, 2026

Giving an AI Assistant a Real Computer: How Notis Runs on Vercel

Why Notis puts the Portal on Vercel and built its first Cloud Computer on Vercel Sandbox (now moving to boat.dev) — snapshots, deadlines, shared caches and the egress problem nobody warns you about.

Reviewed by Florian ParisetAnalysisSep 18, 2026

Reliable AI Agents Need Two Modes: Workflows and Verification Loops

The most reliable AI system is not the most autonomous one. Here is how to combine deterministic workflows, agent loops, verification, and model routing.