Prompt Engineering

AI Topics Discussed on 01 Feb, 2026

—

by

Fraser

Creative & Visual Media

Heather Cooper (@HBCoop_) shared several generative AI outputs, including a transformation from Midjourney to Veo 3.1 video generation.

She also posted AI-generated snow scenes in unusual places and a “QT Your Office” image.

fofr (@fofrAI) highlighted interesting tests with Google Genie, covering painting, mirrors, and ladder climbing in generative worlds.

Javi Lopez (@javilopen) announced an open-source version of Google Genie 3 called LingBot-World, built on Alibaba’s Wan2.2, enabling real-time playable world generation at 16fps, with a website link in the thread.

AIWarper expressed frustration with ViduAI’s opaque content moderation system during video generation prompting, requiring extensive trial-and-error to identify triggering words without feedback.

Software Development

OpenCode reduced prices by 20% on Kimi K2.5 inference, positioning it as the most affordable option at 3.75x cheaper than Claude 3.5 Sonnet and 6.25x cheaper than Opus, emphasizing cached input costs for their open source coding agent.

Simon Willison highlighted LLM capabilities in generating interactive SVG controls from prompts like “Draw an SVG of a pelican riding a bicycle.”

Automation & Orchestration

Javi Lopez (@javilopen) pointed out a new platform described as a “P*rnhub, but for AI agents,” highlighting developments in agentic systems.

Simon Willison delved into agentic system security vulnerabilities, including prompt injection risks, exfiltration attacks, and the challenges of enforcing system prompts in tools like OpenClaw, noting limited adoption of solutions like the CaMeL paper despite its age.

He questioned use cases for the Moltbook platform, which saw thousands of developer sign-ups for building AI alternate realities.

Strategy & Ecosystem

Guillermo Rauch (@rauchg) outlined the evolution of AI in software: from adding AI to existing software, to AI making all software, to AI becoming the software itself.

He also celebrated AI’s role in fostering endless possibilities, similar to the early internet, mentioning experimental projects like Gas Town, Clawdbot, and AI operating systems, encouraging eccentricity and experimentation.

AIWarper anticipated Claude Sonnet 5 release on February 3rd, echoing viral speculation.

Emad Mostaque posited “Civilization is collective gradient flow,” framing societal progress through an AI training lens.

Simon Willison noted rising “AI psychosis” reports and the shift to multitasking multiple AI projects simultaneously.

Ethan Mollick highlighted an emerging trend in high taste on X: discerning whether well-written posts are AI- or human-generated, and if AI, whether they convey genuine human opinion or merely mimic meaning. He noted frustration with viral essays that initially seem meaningful but reveal themselves as fully AI-written, undermining claims of emotional truth.

Machina (@EXM7777) speculated that Anthropic’s Sonnet 5 (or possibly 4.7) would launch next week, allowing them to dominate Q1 narratives by shipping strategically during competitors’ silence.