This New Way

• AI Powers 1500 Developers & Generates Movie Magic with Allan Isfan of Warner Bros

  In this episode, Aydin chats with Allan Isfan, Senior Director of Global Video Platform at Warner Bros Discovery, about how AI is reshaping creativity, software development, and large-scale enterprise culture. Allan explains how he drives AI literacy for 1,500+ employees, the power of internal demos and sandboxes, and gives a hands-on walkthrough of generative video tools like Gemini V3, Flow, and Sora. He also dives into AI video analysis, the Wizard of Oz project at The Sphere, and the future of creative storytelling powered by AI.🕒 Timestamps00:00 – Aydin welcomes Allan Ispan to the show.01:03 – Allan’s career path: Nortel → startups → venture capital → media tech.02:09 – Founding FaveQuest and My Event Apps, powering Ottawa’s festivals.02:53 – Moving to LA with UIV → acquired by WarnerMedia.03:38 – Current role: Senior Director, Global Video Platform for HBO Max, CNN, Discovery+.04:11 – How Warner Bros started its AI journey: three core pillars.04:55 – The vision: “Talk to the app” for content recommendations.05:26 – Scaling AI enablement for 1,500+ employees company-wide.06:13 – Making your company more AI-native without a Head of AI.07:00 – Step 1: Build AI literacy and create role-based learning paths.08:07 – Setting measurable goals: 80% literacy by year-end.08:46 – AI all-hands: excitement, humor, and internal FOMO.09:32 – Launching AI Friday Demos – monthly internal showcases.09:55 – Brown bag sessions for hands-on education (e.g. generative video).10:59 – Internal data querying: using AI on top of Jira and internal docs.11:41 – The origin of AI all-hands → now a recurring company event.12:53 – Experimental budgets, legal review, and security hurdles.14:07 – Creating AI sandboxes for safe experimentation.15:14 – Advice for smaller teams: give employees micro-budgets to experiment.16:54 – Generative video: “State of the art is moving bonkers fast.”17:41 – Demo 1: Google Gemini V3 — 8-second clips from text prompts.18:58 – Prompting tips: scripting short sequences for realism.23:42 – Voice options: when to use Eleven Labs for cloning.26:00 – Advanced camera moves and cinematic continuity.28:06 – “Anyone can be a director now.” Democratizing filmmaking.29:04 – Demo 2: Using Flow to connect multiple AI-generated scenes.33:04 – Cost and quality tradeoffs: fast vs. standard rendering.34:26 – Sora (OpenAI): create cameos and realistic social clips.35:41 – New business models: celebrity likeness + embedded sponsor branding.36:55 – Meta Ads: turning photos into videos for higher engagement.40:04 – Quickplay demo: searching long-form video content (“Smelly Cat” in Friends).42:54 – Live sports AI: tracking, play-by-play, and highlight automation.45:27 – The Wizard of Oz @ The Sphere (Las Vegas) – AI-enhanced 360° remake.47:39 – Allan’s 12-month outlook: expanding creative boundaries with AI.48:33 – Personal note: turning children’s books into AI-animated cartoons.🧰 Tools & Technologies MentionedChatGPT / Claude / Cursor / Windsurf – AI code assistants that boost developer productivity.Gemini V3 & Flow (Google) – Text-to-video generation and multi-scene creation.Sora (OpenAI) – Mobile app for AI cameo video generation.Eleven Labs – Industry-leading AI voice cloning.Quickplay – AI video intelligence and repurposing platform.Adobe Firefly – Generative design and image-to-animation tool.Riverside.fm – Podcast platform with AI-generated highlights.Fellow.ai – AI meeting assistant for notes, actions, and insights.Subscribe at⁠ thisnewway.com⁠ to get the step-by-step playbooks, tools, and workflows.

  --------  

  49:29

  --------

  49:29
• How to Build Vertical AI Businesses Fast with Ryan Carson, Builder in Residence at Sourcegraph

  Ryan Carson (ex-Treehouse, Intel; now Builder-in-Residence at Sourcegraph’s AMP) shares his origin story and a practical playbook for shipping software with AI agents. We cover why “tokens aren’t cheap,” how AMP made pro-level coding free via developer ads, a concrete workflow (PRD → atomic dev tasks → agent execution with self-tests), and why managers should spend time as ICs “managing AI.” We close with advice for raising AI-native kids and a perspective on this moment in tech (think integrated circuit–level shift).Timestamps00:00 – The beginning of intelligence: how LLMs changed Ryan’s view of computing00:23 – Apple IIe → Turbo Pascal → Computer Science: the maker bug bites03:20 – DropSend: early SaaS, Dropbox name clash, first acquisition04:30 – Treehouse: teaching coding without a CS degree; $20M raised, acquired in 202105:02 – The “bigger than a computer” moment: discovering LLMs06:15 – Joining Intel: learning GPUs and the scale of silicon (“my adult internship”)07:09 – Building an AI divorce assistant → joining AMP as Builder-in-Residence09:38 – AMP vs ChatGPT/Claude/Cursor: agentic coding with contextual developer ads11:09 – Token economics: why AI isn’t really cheap17:27 – Frontier vs Flash models (Sonnet 4.5 vs Gemini 2.5) — how costs scale21:31 – Private startup: vertical AI for specialized domains22:36 – The new wave of small, vertical AI businesses23:01 – Live demo: building a news app end-to-end with AMP28:18 – How to plan like a pro: write the PRD before you build30:02 – “Outsource the work, not your thinking.”32:28 – Turning PRDs into atomic tasks (1.0, 1.1…)35:50 – Competing in an AI world = planning well36:28 – Managers should schedule IC time to “manage AI”37:14 – Designing feedback loops so agents can test themselves39:47 – “AI lied to me”: why verifiable tests matter41:11 – Raising AI-native kids: build trust, context, and agency43:59 – “We’re living in the integrated circuit moment of intelligence.”Tools & Technologies MentionedAMP (Sourcegraph) – Agentic coding tool/IDE copilot that plans, edits, and ships code. Now offers a high-end, ad-supported free tier; ads are contextual for developers and don’t influence code outputs.Sourcegraph (Code Search) – Parent company; enterprise code intelligence/search.ChatGPT / Claude – General-purpose LLM assistants commonly used alongside coding agents.Cursor / Windsurf – AI-first code editors that integrate LLMs for completion and refactors.Bolt / Lovable – Text-to-app builders for rapid prototyping from prompts.WhisperFlow / SuperWhisper – Voice-to-text tools for fast prompting and dictation.Anthropic Sonnet 4.5 – Frontier-grade reasoning/coding model; powerful but pricier per token.Google Gemini 2.5 Flash – Fast, lower-cost model; “good enough” for many workloads.Auth0 (example) – Authentication-as-a-service mentioned as a contextual ad use case.GPUs / TPUs – Compute for training/inference; token cost drivers behind AI pricing.PRD + Atomic Tasks Workflow – Ryan’s method: record spec → generate PRD → expand to dot-notated tasks → let the agent implement.Self-testing Scripts – Ask agents to generate runnable tests/health checks and loop until passing to reduce back-and-forth and prevent “it passed” hallucinations.Family ChatGPT Accounts – Tip for raising AI-native kids; teach sourcing, context, and trust calibration.Subscribe at⁠ thisnewway.com⁠ to get the step-by-step playbooks, tools, and workflows.

  --------  

  45:33

  --------

  45:33
• How Designers Become Builders with AI with Filip Skrzesinski of Subframe

  Aydin sits down with Filip Skrzesinski, co-founder of Subframe, to unpack how AI and code-native design tools are collapsing the classic PM → design → engineering handoffs. Filip explains why “pictures to code” is an unfair ask of engineers, shows how Subframe lets teams design directly in the same material as production code, and demos building a Fellow feature—from screenshot → design system match → working prototype—without access to Fellow’s codebase. They close on what’s next: organizations training their own “house models” to reflect product taste, patterns, and constraints so more people across the company can truly build.Key takeawaysDesign in the same material as code: Subframe treats UI work as editable code, eliminating fidelity loss from design handoffs.Fewer stages, faster loops: PMs, designers, and engineers collaborate in one artifact; prototypes look and behave like the real app.AI as a trained teammate, not a slot machine: Teams will shape models with system prompts, snippets, and feedback—like mentoring a junior designer.Front-end ownership shifts: Designers can own front-end structure and components; engineers wire up backends and complex logic.Prototype to PRD: High-fidelity prototypes beat docs for alignment, user testing, and speed.Timestamps00:00 - Introduction 01:00 Fil's path: audio engineering → CS → design → startup co-founder03:48 Builders everywhere: from Dreamweaver → Webflow → Shopify → now “apps”04:01 What Subframe is: a design tool rooted in code05:48 Bridging LLMs (great at code) with visual design context08:09 The architect vs. printer analogy for product design12:23 Back to the show: “The new way” is collapsing steps and handoffs14:07 “Five-year” vision (sooner than you think): design → code with agents in the loop16:31 Training models on your org’s taste: like raising a puppy—examples & theory19:15 Today’s demo plan: build a Fellow feature in Subframe without codebase access21:04 Recreating Fellow’s UI: import colors/typography; screenshot → layout23:07 Don’t fight the AI: let it rough-in, then designers perfect in visual mode24:11 Why prototypes should look native (not “off-brand” sandboxes)26:07 Syncing components to codebases; where Subframe stops (front-end) and engineers continue (backend)28:33 Programmatic (deterministic) UI code & generative for visuals30:00 PMs in the tool: prompt to add a Share dialog with transcript and video context35:08 Exploring multiple design variations; mix-and-match patterns (“snippets”)37:57 From design to interactive prototype via annotations (“do this on click…”)45:22 First build runs: working Share flow; alert updates after sending47:02 Export code → Cursor/GitHub; hand off real components48:08 The next 12 months: more ideas shipped, more makers, less gatekeepingTools & technologies MentionedSubframe — Code-native design tool for building UI/UX; designs directly edit the underlying code; syncs components to your repo.Fellow.ai — AI meeting assistant with privacy controls; accurate summaries, actions, decisions; broad SaaS integrations.Cursor — AI-assisted code editor; good for continuing from exported Subframe code to production.GitHub — Repo hosting and collaboration for shipping the generated/edited UI code.AI code agents — Used by engineers to wire front-end to backend services and data.Squarespace / Webflow / Dreamweaver — Prior waves that democratized web creation; backdrop for today’s “apps layer.”Shopify — Example of no-code/low-code e-commerce; analogy for app building’s democratization.Lovable / Bolts / V0 — AI code/prototyping tools referenced as peers for generating working app scaffolds.Slack / Asana / HubSpot / Salesforce / Linear / Jira / Confluence — Systems Fellow integrates with to push notes, actions, and records.Subscribe at⁠ thisnewway.com⁠ to get the step-by-step playbooks, tools, and workflows.

  --------  

  50:09

  --------

  50:09
• AI + n8n: From YouTube Insights to Sales Funnels in Minutes with JD Fiscus

  JD Fiscus (nerding.io) shares how a late-night hack connecting MCP to n8n exploded to ~1M downloads, then demos practical MCP workflows: indexing YouTube channels for Q&A, and auto-building n8n flows from natural language. We dig into the Agentic Commerce Protocol, real security pitfalls (like destructive commands), and how to turn MCPs into products with OAuth and Stripe for authentication and metered billing. He closes with how he teaches this hands-on at the Vibe Coding Retreat.Timestamps1:00 Why build it: “MCP shouldn’t be Claude-only”—bridging MCP into n8n early (Dec/Jan)2:09 Shipping under the pseudonym nerding.io; surprise seeing creators use it2:25 n8n later ships its own MCP server/client; they nod to nerding.io & Simon3:59 “N8n is useful, but so much more useful with MCP”5:12 What MCP means for software: every smart company is exposing an MCP; new login/usage patterns6:27 Agentic Commerce Protocol (ACP): Stripe + OpenAI; agents checkout across the web8:02 Marketing to agents not humans? SEO shifts as agents comparison-shop9:10 Early “agent mode” attempts vs protocol-based purchases (less hacky)10:58 Likely adopters: platforms (Shopify) & big retailers; echoes of early MCP evolution14:11 Security realities: token passing evolved to OAuth; hallucination + destructive actions risk16:04 Personal mishap: agent ran supabase reset on a dev DB—imagine prod! Guardrails matter17:03 Designing MCP servers: don’t just “wrap your API”; use resources/prompts for agentic UX19:04 Demo 1—Influencer MCP: index a YouTube channel, embed transcripts, ask questions in Claude20:54 Storage: embeddings into Postgres; per-channel tables24:46 Keeping it fresh: daily cron to ingest new videos25:18 Demo 2—Build n8n workflows from chat using N8N MCP (by Ramullet); live docs + API27:00 “Create a webhook → send leads to Sheets” built conversationally, with allow/deny prompts31:02 Zapier, Gumloop: agents that build automations via natural-language steps34:00 Next frontier: custom connectors (Claude/Cursor/OpenAI), OAuth auth flows for MCPs39:03 Turning MCPs into products: login with Twitter → Stripe subscription → metered billing41:12 Paid tool call demo: “paid echo” → Stripe usage event logged per user43:41 How to learn this fast: vibecodingretreat.com (small cohorts, hands-on builds)Tools & Technologies Mentioned (quick guide)MCP (Model Context Protocol) — Standard for connecting models to tools/data; supports tools, resources, prompts.n8n — Open-source automation platform; JD wrote an MCP node that went viral; also has native MCP server/client now.Claude / Cursor / OpenAI (custom connectors) — LLM IDEs/chats that can load MCPs; custom connectors enable OAuth + productized access.Agentic Commerce Protocol (ACP) — Early protocol (Stripe + OpenAI) for agent-initiated purchases with confirmations.Web MCP (W3C-oriented idea) — Emerging patterns for agent↔︎website interactions beyond human UI flows.OAuth — Secure, user-consented authentication for MCPs (vs passing raw tokens).Stripe (subscriptions + metered billing) — Attach billing/usage limits to MCP calls; track per-user consumption.YouTube API + Transcripts — Source data for the “Influencer MCP” indexing pipeline.Embeddings + Postgres — Store vectorized transcript chunks in Postgres for retrieval (JD self-hosts).Cron — Schedules daily ingestion of new content.Google Sheets — Target destination in demo for simple lead funnels.Zapier / Gumloop — Natural-language automation builders; early NLA/agent patterns.Git / CLI commands — Cautionary tale: agents running destructive commands (e.g., resets).Do Browser / Comet Browser — Agentic browsing tools referenced for web actions.Fellow.ai — AI meeting assistant with security-first design; generates precise summaries/action items.Subscribe at⁠ thisnewway.com⁠ to get the step-by-step playbooks, tools, and workflows.

  --------  

  45:52

  --------

  45:52
• Compound Engineering: Manage Teams of AI Agents with Kieran Klaassen of Cora

  Aydin and Kieran Klaassen (Cora) unpack Compound Engineering—treating every task as an investment so the next time is faster. Kieran shares his path from film composer to startup CTO and live-demos how he plans → prototypes → ships a feature using AI agents (Claude Code), then runs multi-agent reviews. They discuss why managers are primed to orchestrate agents, how to capture your own feedback patterns, and why there’s “no excuse not to have a prototype” anymore.Timestamps0:07 — “Every piece of work should be an investment.”2:15 — What Cora is: an AI Gmail layer that auto-archives ~80% and briefs you twice daily.3:32 — Launch notes & early user reactions.5:21 — The Claude Code pricing saga and “finding the limits.”8:06 — Compound Engineering defined (codify how you work so AI does it next time).15:01 — From “automation” to pattern-capturing systems; natural-language rules over brittle workflows. 22:03 — Demo kickoff: planning the “Invite friends” improvement inside Cora.26:11 — Rapid mockups from a screenshot + voice description; iterate in seconds.33:06 — Multi-agent planning: repo research, best-practices scout, framework researcher.41:01 — Human judgment on plans; simplify when encryption/perf add hidden complexity.50:00 — Feature running end-to-end; agentic PR + test flow; sub-agent code reviews.Tools & Technologies MentionedCora — AI inbox copilot for Gmail that prioritizes, summarizes, and drafts replies; batches the rest into twice-daily briefs.Claude Code (Anthropic) — Agentic coding/terminal assistant used for planning, building, and reviews.Monologue — Voice-to-text for quickly describing UI and generating mockups.Every.to — Partner/design/content hub Kieran collaborates with; also publishes his writing on Compound Engineering.GitHub + GitHub CLI — Issues, branches, PRs automated by agents from plan → code → review.VS Code (with Claude Code extension) — IDE setup for hands-on edits when needed.Anthropic Console Prompt Generator — Used to scaffold robust prompts/agents, then refined manually.Model mix for reviews (e.g., “GPT-5 Codecs,” “Claude Opus”) — Alternative model passes for plan/code critique.Fellow.ai — Aydin’s AI meeting assistant for accurate notes, actions, and privacy-aware summaries.Subscribe at⁠ thisnewway.com⁠ to get the step-by-step playbooks, tools, and workflows.

  --------  

  55:03

  --------

  55:03

Más podcasts de Tecnología

Podcasts a la moda de Tecnología

Acerca de This New Way