Prompt Engineering

AI Topics Discussed on 12 Feb, 2026

—

by

Fraser

Creative & Visual Media

Discussions centered around impressive generative video capabilities of Seedance 2.0, with users showcasing short, high-quality clips recreating famous scenes.

Javi Lopez (@javilopen) generated a 15-second Lord of the Rings scene and shared its Spanish version, highlighting the model’s wild potential.

Ilker (@ailker) created a creative Harry Potter video imagining Hermione using a Time-Turner to travel back to 1981, demonstrating advanced video synthesis.

AIWarper demonstrated Seedance 2.0’s capabilities by transforming static frames from the manga *Vagabond*—never before adapted into anime—into a fluid animated sequence, highlighting its potential to revolutionize content creation for manga artists.

AIWarper also shared an impressive AI-generated video clip, noting rapid desensitization to advancing video quality within months.

Discussions highlighted impressive advances in AI video generation. Ilker (@ailker) shared a 15-second Lord of the Rings recap using Seedance 2.0, calling it incredible, which garnered massive engagement.

He also demonstrated a creative Hermione Granger time-travel video using a Time-Turner to 1981.

Justine Moore (@venturetwins) spotlighted an AI-generated adult version of Sesame Street and celebrated KREA AI’s acquisition of Wand app with a new iPad app featuring real-time AI brushes.

Ethan Mollick (@emollick) showcased rapid progress in generative textures for GTA mods using Nano Banana Pro, deployable in-game via Claude skills.

Software Development

Javi Lopez (@javilopen) highlighted Google DeepMind’s upgrade to Gemini 3 Deep Think, a specialized reasoning mode for tackling science, research, and engineering challenges, including semiconductor material design as shown in a Duke University lab demo.

Simon Willison tested OpenAI’s new GPT-5.3-Codex-Spark model, noting its faster performance for building applications but slightly inferior visual outputs compared to GPT-5.3-Codex in pelican generation tests.

He linked this to his blog analysis and observed no holiday dip in coding activity at Vercel, attributing it to AI model experimentation.

Willison also commented on Anthropic’s Claude Code revenue doubling to $2.5B run-rate since early 2026 amid surging user growth.

POM (@peteromallet) introduced Desloppify, an agent toolset for beautifying “slop code,” reporting strong results in his projects and inviting testing.

Automation & Orchestration

Jonathan Fischoff (@jfischoff, AI at fal) predicted future job categories: agent managers overseeing AI systems, regulated button-pushers for compliance-critical tasks, and frame moggers curating video frames.

OpenRouter (@OpenRouterAI) launched MiniMax M2.5, an upgraded agentic model excelling in reliability for long-running tasks beyond code writing.

Desloppify was noted for its agent-driven code improvement workflows.

Strategy & Ecosystem

@levelsio questioned Microsoft’s absence from top LLM benchmarks despite its OpenAI stake, urging them to train proprietary models.

He advocated for U.S. AI firms going public (e.g., Google, upcoming OpenAI/Anthropic IPOs) to enable broad investor participation in automation-driven growth, contrasting with opaque Chinese markets.

Ethan Mollick emphasized frontier AI’s massive scale and growth, citing Anthropic’s explosive metrics as evidence against bubble concerns.

He also observed near-saturation of ARC-AGI benchmarks in under a year, signaling fast progress in reasoning capabilities.

Machina (@EXM7777) argued AI creation mirrors human artistry through recombination at superhuman speed and warned against overhyping early ARR in AI agencies.