Divya van Mahajan

10+ Things You Should Build With AI Instead of Sending Files

ai-daily-brief-podcast

Overview

The talk argues that the simple, AI-generated website or web app is becoming a superior replacement for many traditional knowledge-work file formats (decks, PDFs, spreadsheets, memos, etc.), and that the convergence of AI coding tools and simplified publishing platforms makes this shift accessible to any knowledge worker today. The speaker is the host of the AI Daily Brief podcast and video channel. The episode was prompted by OpenAI’s launch of Codex Sites, a feature that simplifies publishing Codex-built projects as shareable websites.

Source video URL: (not available)

Prerequisites

  • Basic familiarity with knowledge-worker productivity tools: slide decks, PDFs, spreadsheets, shared documents, email workflows.
  • General awareness of AI-assisted coding tools (e.g., OpenAI Codex, Claude, Replit, Lovable, Gamma) and what “vibe coding” means in practice.
  • Awareness of web hosting concepts (URLs, static vs. dynamic sites) at a conceptual level — no coding skill required to follow the argument.
  • Optional: familiarity with tools like Vercel, Supabase, Docsend, Google Docs, and Notion as context for the problem space.

Main Points

OpenAI Codex Sites as the Catalyst

  • OpenAI announced updates to Codex including a Sites feature that enables users to publish built projects as accessible websites or web apps without separately wiring together hosting and database platforms.
  • Previously, publishing required combining tools like Vercel (hosting), Supabase (database), and an AI coding agent, or using all-in-one platforms like Replit or Lovable.
  • Codex Sites brings that all-in-one publishing experience natively into Codex, lowering the barrier significantly.
  • This launch crystallises a broader trend: the website as a new default unit of knowledge-work output.

The Core Problems with Traditional File-Based Artifacts

The speaker identifies ten or more structural problems with downloadable files that websites inherently solve:

  1. Versioning / currency — Downloaded files are snapshots; a URL always serves the latest version, eliminating plan_final_v8 proliferation.
  2. Distribution friction — Files require compatible software, download bandwidth, and re-forwarding; a link works universally across email, Slack, SMS, CRM, phone, or laptop.
  3. Navigation constraints — Docs are linear, spreadsheets are tabular, PDFs are paged; websites can organise the same material by topic, role, urgency, or depth, letting each reader self-direct.
  4. Context fragmentation — Knowledge is typically scattered across spreadsheets, memos, browser tabs, and email threads; a single URL can layer all of that context together.
  5. Passive consumption — Traditional documents are one-way; websites enable interactive elements (toggles, filters, embedded media, forms).
  6. Audience fit — Files force an “average reader” assumption; websites can branch by audience type or role.
  7. Actionability — Documents point to actions that must happen elsewhere; websites can embed the action layer natively.
  8. Reusability / modularity — Web content is composable: sections link out, charts embed elsewhere, private versions become public ones, proposals become portals.
  9. Observability / feedback — Files go into a void; websites can generate analytics on what was read, clicked, searched, shared, or abandoned, making the artifact improvable.
  10. Presentation quality — With AI tools, websites can look polished and intentional; the equivalent of a well-designed deck but with far more flexibility.
  11. Permissioning — Websites support granular access control for different audience segments.
  12. Agent-readiness — Cloudflare reported that bot/agent browsing now exceeds human browsing; HTML-based artifacts are structurally more compatible with agent consumption than PDFs, CSVs, or PowerPoints.

18 Specific Use Cases for Website-First Knowledge Work

The speaker walks through eighteen examples of traditional artifacts that are better built as websites:

  1. Slide deck → Narrative website — Breaks free from 16:9 constraints, enables linking, and is already being approximated by tools like Gamma.
  2. Strategy memo → Strategy site — Layers context, argument, evidence, and objections in a navigable, audience-aware structure rather than a linear wall of text.
  3. Research report → Research hub — Makes dense findings accessible, layered, and linkable across a wide range of readers.
  4. Spreadsheet → Data site — Transforms a creator-centric tool into a guided view with dashboards, filters, and visualisations for any audience.
  5. Sales proposal → Proposal microsite — Enables interactive pricing/ROI toggles and provides observability into prospect engagement.
  6. Client update → Client portal — Consolidates status, milestones, deliverables, and open questions into a single living URL instead of scattered emails and decks.
  7. Project brief → Project homepage — Maintains a canonical, evolving reference for goals, stakeholders, and decisions throughout a project’s life.
  8. Case study → Interactive case page — Preserves the richness of a story (problem, process, metrics, video) that a PDF necessarily compresses.
  9. Competitive analysis → Competitive intelligence hub — Transforms a one-time snapshot into a living resource, with future potential for agents to continuously update it.
  10. Static training materials → Dynamic learning site — Purpose-built learning experiences (cited: Agent OS, Claw Camp, AIDB New Year program) whose cost to produce has collapsed due to AI tools.
  11. Employee handbook → Living handbook site — Turns a document nobody reads into an always-current, interactive resource.
  12. Board materials → Board portal — Separates required reading from backup material; provides context without burying the key information.
  13. Investor update → Investor page — Provides always-current (potentially real-time via APIs) metrics, reducing one-off information requests from investors.
  14. Recruiting packet / job description → Candidate site — Gives a richer, more navigable picture of the role and company to help candidates self-qualify.
  15. Brand guidelines PDF → Brand system site — Single URL with all current assets and guidelines, replacing folder-based PDF hunting.
  16. Media kit → Press site — One link with everything press contacts need, always up to date.

(The speaker references 18 examples but the transcript’s usable content covers the above set.)

The Vibe Coding Macro-Trend

  • The speaker characterises much of the explosion in “vibe coding” as knowledge workers discovering that websites are better information-sharing vehicles than traditional file artifacts.
  • Platforms embedding publishing features (e.g., Codex Sites) will accelerate this adoption further.
  • The long-run implication is that in an agentic paradigm — where AI agents browse and consume content — the HTML web is structurally more durable than legacy file formats.

Key Concepts

  • Codex Sites — A feature in OpenAI’s Codex that allows users to publish built projects directly as shareable websites or web apps, without needing separate hosting or database infrastructure.
  • Vibe coding — The practice of using AI tools to generate functional software (including websites) through natural-language prompting, without traditional programming expertise.
  • Knowledge-work artifact — Any packaged output of professional intellectual work: decks, memos, reports, spreadsheets, proposals, handbooks, etc.
  • Anchor artifact — The primary deliverable or container format in which a knowledge worker packages and shares thinking (e.g., the deck used to be the anchor artifact; the argument is the website is becoming one).
  • Proposal microsite — A bespoke, interactive web page built to replace a static sales proposal, enabling dynamic pricing/ROI exploration and prospect-engagement analytics.
  • Competitive intelligence hub — A living website version of a competitive analysis, designed to be continuously updated (potentially by AI agents) rather than published once.
  • Agent-readiness — The degree to which a digital artifact is structured for consumption and processing by AI agents; HTML and web standards score significantly higher than PDF or proprietary file formats.
  • Observability — The capacity of an artifact to generate data about how it is being used (reads, clicks, searches, drop-offs), enabling iterative improvement.
  • Permissioning — Access control mechanisms that determine which users or audiences can view which parts of a site or document.
  • Gamma — An AI-native presentation tool cited as an example of the collapsing boundary between slide decks and websites.
  • Docsend — A PDF distribution platform cited as an intermediate solution to the versioning/currency problem that websites solve more generically.
  • Vercel / Supabase — Hosting and database platforms respectively, previously required to publish AI-built projects before all-in-one tools like Codex Sites existed.

Summary

The speaker’s central argument is that the website — now trivially producible by any knowledge worker using AI coding tools — is a structurally superior replacement for most traditional file-based work artifacts. Where files are static snapshots burdened by versioning, distribution friction, linear navigation, and passive consumption, websites offer living, linkable, interactive, observable, and agent-ready alternatives. Triggered by OpenAI’s launch of Codex Sites, the episode catalogues roughly a dozen systemic problems with legacy document formats and then maps eighteen concrete artifact types — from slide decks and strategy memos to investor updates and employee handbooks — where building a website is either a meaningful improvement or a fundamental transformation of the artifact itself. The speaker frames the broader surge in vibe coding as, in large part, knowledge workers organically discovering this shift, and predicts that as AI platforms continue to embed one-click publishing features, the website will increasingly become the default unit of professional knowledge output.