Prompt Engineering

AI Topics Discussed on 22 Feb, 2026

—

by

Fraser

AI Topics Discussed on 22 Feb, 2026

Creative & Visual Media

Heather Cooper (@hbcoop_) showcased progress in generative AI video models, focusing on achieving smooth, realistic motion and intricate details in image-to-video generation. She shared a clip of a young woman in a hospital setting, emphasizing subtle jaw tension, thumb pressure, lighting effects, and ambient sounds like footsteps and announcements.

Justine Moore showcased how real estate is leading in AI video adoption, using generative tools to animate properties for social media-style ads and visualize potential developments like building options on land plots.

Ethan Mollick pointed out that AI’s capacity to process real-time video and images remains largely underexplored despite valuable applications.

Software Development

@levelsio shared his experience hitting rate limits on Claude Code’s $100/mo Max plan while rapidly fixing the last 10 bugs and feature requests for ideasandbugs.com by copy-pasting issues into the tool.

He argued that for low-to-mid-level development work, Claude Code outperforms human devs by being faster, more reliable (no sleep or sickness), and cheaper, preserving value only for top-tier engineers or those leading AI effectively.

@simonw highlighted how AI coding agents are eroding SaaS moats built on implementation complexity, citing SAML integration dropping from months to days, enabling faster feature iteration despite persistent challenges like market fit and trust.

Justine Moore demonstrated vibecoding a CRM using Claude, highlighting AI-assisted full-stack development workflows.

Automation & Orchestration

Machina shared techniques for 100x improving AI outputs by spawning parallel sub-agents in models like Opus 4.6 to critique drafts using real expert frameworks extracted via NotebookLM.

He also outlined building an autonomous SEO agent in OpenClaw with programmatic page generation, keyword research via APIs like Keywords Everywhere and DataForSEO, and cost optimizations.

Additionally, Machina provided a detailed prompt for OpenClaw to audit personal workflows, prioritize bottlenecks, and deploy sub-agents for high-impact automation.

The Boring Marketer declared the traditional media buyer/ads manager role obsolete due to AI automation.

Strategy & Ecosystem

@EMostaque noted AI labs face pricing friction for high-token-use products like constant Claude Code runs, but emphasized tokens are dropping 10x cheaper annually (potentially 100x), easing constraints.

He quipped on the “throw more tokens at the problem” rule with a book promo image.

Ethan Mollick critiqued overestimation of rapid corporate AI adoption, citing company inertia and AI’s jagged capabilities that slow system-level integration despite fast task-level changes.

He warned that AI lab CEOs’ ominous job loss rhetoric risks policy backlash as AI gains broader attention, urging clearer positive visions.

Mollick highlighted benchmark saturation where weaker LLM judges fail to evaluate stronger models, positioning judges as the new bottleneck.

Machina announced ambitions to build and scale the greatest AI community to $100k/month revenue in 2026.