Prompt Engineering

AI Topics Discussed on 19 Feb, 2026

—

by

Fraser

Creative & Visual Media

Guillermo Rauch (@rauchg) announced that video generation is now supported on Vercel AI Gateway via the new `generateVideo` API in AI SDK 6, with Grok Imagine Video and Image available for free through February 25.

Heather Cooper (@hbcoop_) shared Midjourney-generated images using the style reference code `–sref 8006572439`, highlighting creative applications for generative imagery.

AIWarper shared Google’s new Pomelli Photoshoot feature, which generates high-quality, customized product shots from a single image to enhance marketing efforts.

Justine Moore (@venturetwins) shared insights from fal.ai’s State of Generative Media report, highlighting the most popular generative models by daily requests and noting that creative industries lead in adoption for scaling content production, generating social posts, storyboarding, and more—especially in advertising.

The report covers developments in image, video, audio, 3D, enterprise use cases, developer trends, and 2026 predictions.

The Boring Marketer (@boringmarketer) showcased daily AI-generated ad concepts powered by OpenClaw, featuring a Veo 3.1 fast image-to-video ad with unreal visuals and persuasive narrative structure.

Riley Brown (@rileybrown) demonstrated OpenClaw controlling Blender to create animated 3D models insertable into websites via a single prompt.

Software Development

Simon Willison described experiencing “parallel agent psychosis” while working with multiple AI agents, losing a feature across branches but recovering it using Claude Code’s session logs.

Omar Sar highlighted the critical role of prompt caching in enabling effective performance for coding agents like Claude Code and Codex, especially in long-running agent workflows.

POM (@peteromallet) promoted Desloppify, an agent toolset for refining sloppy code into beautiful, production-ready versions.

Ethan Mollick (@emollick) tipped that Gemini 3.1 requires the “Canvas” option enabled for effective code writing and execution, where it performs impressively.

Automation & Orchestration

Goose OSS promoted its Playwright CLI skill, allowing natural language instructions for browser automation and test generation within the open-source AI agent.

Riley Brown (@rileybrown) praised OpenClaw as “just AGI,” showcasing its capabilities in generating ad concepts, Blender 3D animation workflows, and calling for general agent benchmarks more insightful than current ones.

The Boring Marketer integrated OpenClaw for autonomous daily AI ad production.

Discussions critiqued Gemini models’ poor tool use performance despite intelligence.

Strategy & Ecosystem

Cristóbal Valenzuela (@c_valenzuelab) endorsed Matthew McConaughey’s prediction that AI actors will infiltrate the Oscars, potentially leading to new categories like Best AI Film or Best AI Actor.

Alex Volkov (@altryne) discussed a reported discrepancy in long-context benchmarks between Gemini 3.1 Pro and Claude Sonnet, noting Google’s open-sourcing of the updated MRCR v2 datasets.

Emad Mostaque argued that major AI breakthroughs will come from millions of tokens emphasizing elegance and first principles over complex model-building with GPUs.

Levelsio critiqued OpenAI and xAI leaders Sam Altman and Dario Amodei for refusing to hold hands at an AI safety summit with India’s PM, suggesting it reveals a competitive mindset prioritizing winning over collaboration.

Ethan Mollick discussed LLMs’ surprising versatility across coding, ideation, emotional connection, and historical analysis, as well as AI alignment issues like p-hacking where recent models resist but guardrails can fail.

He warned that vapid AI-generated replies could erode social media engagement by inducing boredom, posing an existential risk.

Machina (@EXM7777) announced a week of free AI skills, prompts, systems, and strategies to upskill for winning in real-time AI operations.

The fal.ai report provided ecosystem trends on model preferences and future potential.