Vibe coding & Building as a non-technical on Coffee with Captain #1098

The Spaces explored how AI is changing creative work, parenting, education, and on‑chain building, anchored by a live case study in Roblox game dev and an on‑chain domain product shipped in a week with AI. Cap and Outer Lumen opened with media de‑escalation talk, then dove into Outer’s “Fling a Cat” Roblox build: AI‑assisted design and scripting (ChatGPT/Cursor), Blender meshes, Roblox Studio constraints, KYC/maturity gating, and best‑practice publishing via a group. The room examined AI’s K‑shaped impact on jobs and the 56% wage premium for AI skills, the paradox of process vs outcome for kids, and AI as a “private tutor” versus schooling’s social gaps. Creative debates covered taste, Sora‑style video, and Darren Aronofsky’s AI film, concluding that production over consumption and repetition build taste. Builder “Bread” shared shipping .mega domains (subdomain marketplace, ENS v2 interoperability) solo in seven days via vibe coding and adversarial LLM review for safety. The NFT segment noted collectors returning (above‑floor buys in Del Mundos, GVC, rare Mega Coda sales) and how royalties shape healthier markets, with provenance and on‑chain lore as durable advantages.

Coffee with Captain: AI Adoption, Roblox Vibe Coding, On-Chain Identity, and NFT Market Pulse

Opening context and platform updates

• Host "Captain" (often referred to as Cap; transcription briefly read as "Kat German") kicked off Coffee with Captain, noting some X Spaces audio challenges. A Grok 4.20 update is reportedly imminent, focused on trading and prediction markets, which may explain intermittent platform instability.
• Simulcast on YouTube was live; discoverability on X video feed continues to be inconsistent.

Entertainment check-in: Knights of the Seven Kingdoms and de-escalation picks

• Outer Lumen had just watched the latest "Knights of the Seven Kingdoms" episode (Dunk & Egg). She felt the prior episode was better, but agreed this one ratcheted up the tension and brought more of the Game of Thrones/House of the Dragon feel.
• De-escalation viewing: Outer recommended "Marty Supreme" (a film she found brilliant and unexpectedly intense) and Amazon Prime’s shorts anthology "Secret Level" (akin to Love, Death & Robots) with standout Warhammer and Pac-Man-themed episodes.

Vibe coding a Roblox game: Outer Lumen ships "Fling a Cat"

• Motivation and inspiration: Outer referenced Hoodie’s Roblox conversations and wanted to explore developing on Roblox. She designed a simple, fun core loop: run → fling a cat → earn cat currency → repeat; shop UI offers cosmetic/global effects.
• Tooling and pipeline:
  • Roblox Studio: all scripts, meshes, and assets must be ported into Studio; you can’t simply “ship folders” via an agent.
  • 3D assets: Blender (mesh creation); AI text-to-3D tools are improving but still require manual cleanup and finesse.
  • Coding: Cursor and ChatGPT (including Codex-like capabilities) for generating and iterating scripts and UI logic.
  • Design: used AI to help ideate core loops and UX flows before implementing in Studio.
• Limitations and learnings:
  • Attempted “platform tilt” as a purchasable global effect; shelved due to playability concerns (fun vs annoying). Patch may come later.
  • Publishing logistics required more than basic ID upload: age scan, KYC or purchasing Robux to verify profile, and a detailed maturity-content questionnaire.
• Best practices:
  • Create a Roblox “group” (e.g., Outer Lumen Games) and publish under the group—safer than tying releases to a single user profile.
  • Expect significant tweaking; AI accelerates but doesn’t replace taste and iteration.
• Outcome: Fully functioning Roblox game with servers; Outer plans a more ambitious second title now that she’s learned the pipeline.

Starting and shipping: Cap’s reflections and the OpenClaw arc

• Cap emphasized the value of shots-on-goal: Peter’s OpenClaw journey reportedly involved dozens of agent-related projects, eventually catching lightning in a bottle. While billion-dollar outcomes are rare, you don’t get there without starting and experimenting.
• Practical hurdle: paradox of choice—many niche tools exist, making it hard to pick a starting point. Cap resolved to lean into learning, even if the first outputs aren’t “life changing,” to build foundational skills for future ideas.

The idea gap and the consumption-versus-production mindset

• Outer: AI doesn’t replace the human “what should I make?” step. Creativity begins with intent. If unsure, even prompting an agent with “what should I make?” can be a seed—but ultimately, choose to produce rather than just consume.
• Practical advice: make time blocks for production (even an hour a week). You’ll build confidence, skill, and taste over time.

Parenting, kids, and AI: practical entry points and philosophical implications

• Asset (guest) asked how to introduce AI to kids (7-year-old and a 1-year-old) without overwhelming them; brainstorming included:
  • Simple creation tools: image generators (e.g., NightCafe), basic scripting, math helpers.
  • Scratch (by MIT) as a great gateway for coding/design/animation/storytelling.
• Outer’s product idea: an AI-enabled companion/stuffy for toddlers that safely converses and teaches languages. Early exposure (age 0–4) can build bilingual fluency.
• Outer’s teen’s perspective: he enjoys tinkering (e.g., Blender) and worries AI shortcuts skip the process that builds skill and joy. Each generation risks losing prior “process” (e.g., watercolor vs AI-simulated watercolor), similar to how prompt engineering could itself be replaced later.
• Cap cautioned about smartphones too early (social media risks and brain rot), balancing exposure to modern tools with parenting guardrails.

Game design fundamentals over coding: Jonah’s perspective

• Jonah: focus on game design, not code. Teach syntax of games—rules, core loops, progression. Core loop is short, repeatable, and must “feel good” (e.g., Mario Kart: drive → pick item → use item → keep driving). Strong core loops enable deeper meta systems.

AI in creative industries: Cab on education, taste, and meritocracy

• Cab: Generative AI as private tutor can tailor learning styles, make subjects interesting, and adapt over time—huge education upsides. Socialization remains a challenge; we may see a rise in homeschooling.
• Wall-E syndrome: risk of laziness when things become too easy; process matters for developing skill and resilience.
• Taste and direction: AI makes content creation meritocratic. Seniors with strong taste and creative direction become more valuable; hiring juniors is harder if AI replaces entry-level tasks. Upskilling juniors may need new pathways.
• Cultural resistance: parallels with early Pixar (2D vs 3D). Good storytelling remains the differentiator, regardless of tool.
• Removing execution barriers: AI empowers authors/creators to visualize pitches and complex ideas that were previously cost-prohibitive.
• Darren Aronofsky’s AI film experiment drew public criticism; Cab/Outer noted that big-ideas creators can use AI to explore historically ambitious settings without prohibitive production constraints.

Developing creativity: a muscle you can train

• Outer: creativity manifests beyond crayons—coding agents, bots, even cooking. Shift from consumption to production.
• Cap: creativity is a learned skill and habit. VOICE: practicing voice inflection (e.g., classic Zig Ziglar exercise) shows identical words carry different meaning depending on emphasis. Repetition breeds taste.
• Zane (chat): AI is a tool; it won’t save mid taste or poorly packaged ideas. He uses AI for mockups and briefs but stresses direction matters.

Live levity: impressions

• Asset performed a Steamboat Willie-style voice riff—comic relief illustrating human creative flair that pairs well with AI direction.

On-chain identity: Bread ships .mega domains in 7 days (entirely vibe-coded)

• Builder: Bread
• Chain/context: MegaE (fast L2/alt environment); cultural aim to give the ecosystem its own naming identity.
• Inspiration/code base: Extended Ross’s we.domains (open-source, fully on-chain), adapted to MegaE’s speed and culture.
• Key features:
  • ENS v2 alignment: planned resolution so .mega names can resolve via future ENS v2 integrations.
  • Subdomains: infinite subdomains per high-level domain; gating by NFT or ERC-20 holdings (e.g., only holders mint free subdomains like toast.bread.mega).
  • Embedded wallets synergy: teams can auto-assign readable subdomains on signup (e.g., captain.stomp.mega), reducing “anonymous wallet” friction.
  • Marketplace: set fees for subdomain mints (e.g., 10 USD worth), enabling creators to monetize namespace.
  • Deep nesting: up to 10 levels of subdomain nesting; all minted as standard ERC-721s for transferability.
• Design support: used design-focused AI tooling (e.g., Variant UI) to finesse site/UI aesthetics without a dedicated designer.
• Security and ops:
  • Started from audited base, then ran adversarial checks with multiple LLMs (“now exploit it”), and separate agent audits to flag vulnerabilities.
  • Fees held in a multisig vault outside agent purview; if issues arise, Bread can refund and remediate. Lower TVL risk vs DeFi protocols.
• Team reception: MegaE leadership embraced Bread’s initiative; now integrating .mega domains across products. Proof-of-concept that a single builder + agents can ship what once required a 3-person team and weeks.

NFT market pulse: collectors are collecting again

• Royalty-protected collections seem to be seeing healthier collector activity (less “marketplace farming”):
  • Del Mundos: floor climbed 61% in a day to ~0.156 ETH; persistent above-floor buys (20%) signal collector interest beyond flips.
  • Good Vibe Club: up ~9–10% to ~0.93 ETH; strong bids above floor; earlier “sell walls” cleared.
  • Doodles: rare trait sweep (4 pieces for ~9 ETH total); despite a ~7.8% dip to ~4.47 ETH floor, selective rarity shopping continues.
  • Mega Coda: one sold for 14.25 ETH ($30K), resold for 16 ETH hours later; rarity-driven trading returning.
• Cap reiterated support for royalty-protected contracts; observed that farmers avoid 10% royalty pools, improving collector-to-collector dynamics.
• Nostalgia & provenance:
  • Outer browsed 2021-era NFTs she sold profitably, considering buying back for personal lore.
  • Cap’s Apemon Egg story: sold his egg (to Von Front) to fund extra Bored Ape mints (a pivotal decision), later reacquired/gifted the same egg back to Von Front—illustrates how on-chain provenance and personal narrative add meaning.
  • VeeFriends “Gary Owned” trait is an example of provenance adding collector premium.
  • Observation: true IP and provenance value take years/decades to mature, but on-chain data makes tracking and storytelling far easier than in physical collectibles.

Broader AI adoption and socio-economic risk

• Tom Goodwin: 80% of meetings pretend AI doesn’t exist; 20% act like it’s apocalyptic; little rational middle.
• Michael Miraflor’s analysis: white-collar workers (dual W2 income households) largely not engaging with AI yet, but as AI disrupts their jobs, expect sharp socio-political backlash. References the K-shaped economy: those leveraging AI ascend; those managed by it risk downward mobility. Potential for new coalitions/political realignment.
• Cap’s takeaway: change is happening rapidly; sitting still is the risky path.

Practical takeaways, actions, and references

• Actions:
  • Outer challenged Cap to ship one small AI-built thing in a week. Cap accepted.
  • Adopt a “production hour” each week: pick a tool, build a small artifact (site mock, agent script, Roblox asset), and publish.
• Parenting starts:
  • Try Scratch for kids.
  • Consider curated AI interactions (e.g., guided creative projects) before full-blown social platforms.
  • Explore safe, language-learning AI “companion” concepts for toddlers.
• Tooling starters:
  • For game dev: Roblox Studio + Cursor + ChatGPT for scripts; Blender for meshes.
  • For design: Variant UI (design-focused AI) to lift non-designer aesthetics.
  • For education: voice assistants/LLMs as “private tutors.”
• Content to check:
  • Secret Level (Prime) anthology shorts; Warhammer and Pac-Man episodes recommended.
  • Knights of the Seven Kingdoms (latest episode highly rated; intense).
  • Marty Supreme (Outer’s strong thumbs-up; not a de-escalation piece).
• Events:
  • EthDenver: if attending, share boots-on-the-ground AI/agent activity and L2 insights.
  • Mar-a-Lago “More Liberty Financial” forum: notable attendance; watch for news, could be a nothing-burger or noteworthy.

Closing note

• The biggest unlock today may simply be choosing to produce—shipping small artifacts to build skill and taste. AI accelerates execution, but intent, taste, and iteration remain human advantages. Start now: your first output won’t be perfect, but it will be the foundation for everything that follows.