Skip to content
a16za16z

OpenClaw, Claude Code, and the Future of Software | Peter Yang on The a16z Show

Anish Acharya speaks with Peter Yang, creator and product lead at Roblox, about how personal AI agents are replacing the apps we open every day, why coding agents feel like slot machines, and what happens when the cost of building software drops to near zero. They discuss why future companies will stay radically small, how the IDE is becoming a thinking tool rather than a making tool, and why human ambition will always create more jobs than AI eliminates. Timestamps: 0:00—Intro 1:56—Using OpenClaw for voice, memory & daily life 6:14—Will agents kill apps & SaaS? 11:57—Coding agents: Claude Code vs. Codex 17:00—Future of work: small teams, agents & company culture 24:00—How agents change consumer products & the economy Read the full transcript here: https://www.a16z.news/s/podcast Resources: Follow Peter Yang on X: https://x.com/petergyang Follow Anish Acharya on X: https://x.com/illscience Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: https://twitter.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX Listen to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711 Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures.

Peter YangguestAnish Acharyahost
Apr 5, 202629mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

How OpenClaw and coding agents reshape apps, work, and products

  1. Peter Yang describes using OpenClaw mainly as a personable, voice-first “companion” that also performs real tasks like pulling analytics, updating docs, and even initiating Twilio phone calls.
  2. They argue task-oriented apps and simple SaaS are most vulnerable to agent substitution, while entertainment and “feeling-based” apps may be more resilient and apps may evolve into dual UI+API products.
  3. The conversation contrasts Claude Code’s customizable, “chatty/slot-machine” workflow with Codex’s slower-but-more-accurate style, emphasizing different tools for “vibing” vs. serious execution.
  4. They predict smaller, higher-leverage teams aided by agents, with fewer coordination rituals (OKRs/meetings) and less emotional friction in cross-functional negotiation.
  5. They explore how consumer business models may simplify as AI introduces real marginal costs (inference) and makes subscriptions/usage-based pricing more viable, while an “agent stack” (identity, payments, marketing, MCP/CLI) replaces older playbooks.

IDEAS WORTH REMEMBERING

5 ideas

For many users, “agent-ness” is mostly an interface and relationship shift.

Yang says 70–80% of OpenClaw’s value is its personable, always-available Telegram + voice presence, which makes it feel more like texting a human than opening ChatGPT.

Agent memory is still brittle, and users must engineer around it.

OpenClaw’s default file-based memory (daily memory.md) “forgets,” so Yang adds a multi-layer memory/search setup and even instructs the agent (agents.md) to check memory before answering.

Task apps are more exposed to disruption than entertainment apps.

Yang reports he opens apps like banking/analytics less because he can text the agent; they hypothesize entertainment and emotion-driven usage (TikTok, social feeds) will persist longer.

Single-agent UIs create an intent-collision problem that apps previously solved.

Acharya notes apps partition intent (“feel connected” vs. “get work done”); Yang compensates by creating multiple Telegram channels (separate threads) to mimic contextual separation.

Claude Code and Codex map to different working modes: flow vs. rigor.

Yang uses Codex “when I want to do something real” (more accurate, thinks hard, slower) and Claude Code for “vibing” (faster, more variable outcomes, more engaging UX).

WORDS WORTH SAVING

5 quotes

And since that software will eat the world, I, I feel like coding will eat all knowledge work, right?

— Peter Yang

But if I was honest with you, dude, like I mostly just talk to it through voice and get voice replies, and like every other day I ask it to give me like a pep, pep talk.

— Peter Yang

Like it was some-something about like, uh, like, oh, you're like talking to me about your creator business and blah, blah, and like your, your job, but, like, just remember that your kids, you know, seven and four, are gonna grow up very soon, and they're gonna wanna spend time with you.

— Peter Yang

But like I feel like as a company gets bigger, it tends to get shit... It tends to become like a shittier, shittier place to work, dude.

— Peter Yang

But like that, that's the way I work now. I, I never start from zero. Like I always get the first 80% from AI.

— Peter Yang

OpenClaw as Telegram-based personal agentVoice interface, personalization, and “pep talk” use casesMemory systems (file-based memory.md, multi-layer search) and reliability gapsAgents vs. apps: task completion, retention, and context switchingClaude Code vs. Codex: UX, accuracy, customization, and flow stateVibe coding replacing internal SaaS and limits of DIY maintenanceFuture of work: small teams, fewer meetings, negotiation via agentsConsumer monetization shifts: subscriptions + token/usage costsEmerging agent stack: identity, payments, marketing, MCP/CLIAgents and entrepreneurship/solopreneurs, education implications

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome