Prompt Engineering

AI Topics Discussed on 01 Mar, 2026

—

by

Fraser

AI Topics Discussed on 01 Mar, 2026

Creative & Visual Media

Heather Cooper (@HBCoop_) shared a Midjourney style reference (–sref 5463041427) for creating fun generative images, accompanied by a video example.

Javi Lopez (@javilopen) highlighted issues with Seedance 2.0, noting that the generative video model is being nerfed with stricter content filters blocking prompts like “Pornhub but with shotguns and explosions,” leading to frustration over increasing censorship in content creation tools.

Justine Moore (@venturetwins) praised an innovative AI video series turning history lessons into immersive experiences, allowing viewers to walk through historic scenes with a guide, with the creator planning a full series.

POM (@peteromallet) launched the Arca Gidan Prize, a $50k open source AI art competition themed “Time” with sub-themes like Déjà Vu, The Briefness of Bloom, and Traveling Through Time, aiming to inspire novel applications of open models beyond typical trailers or videos.

Software Development

Developers are experimenting with AI configuration files such as AGENTS.md, CLAUDE.md, and Copilot instructions in open-source repositories, but adoption remains low at just 5% across 10,000 scanned repos, with many files static after initial creation and lacking standardization.

@levelsio praised Claude’s Remote Control feature for enabling seamless code editing on production servers from MacOS, iOS, or the web, sharing a custom “yolo mode” command alias: `c() { IS_SANDBOX=1 claude rc –dangerously-skip-permissions “$@”; }`.

Simple PHP/JS stacks without complex builds were highlighted as particularly effective for AI-assisted development due to their straightforward nature.

The “Clawconomy” around OpenClaw (likely an AI coding tool) was noted, with 125 startups generating $274k in verified revenue last month, suggesting significant market impact if representing 1% of the space.

POM (@peteromallet) celebrated two of his open source repositories reaching 1k GitHub stars after 3.5 years of work in AI tools.

Automation & Orchestration

Ideas for autonomous agents that monitor error logs, feature requests, and bug boards to automatically generate pull requests were discussed, with safety considerations like isolating the agent on a separate VPS with read-only repo access to prevent prompt injection.

Riley Brown (@rileybrown), cofounder of VibeCode, shared insights from 200 hours testing OpenClaw: keep agents narrow and focused with clear intent, building teams of specialized agents rather than overloaded generalists, demonstrated via YouTube agent examples.

He emphasized shifting from prompts to giving agents purpose, skills, integrations, tasks, and personalities.

Strategy & Ecosystem

AI’s potential to govern or dictate terms to governments was debated, with predictions that AI will inevitably run governments and questions about control and sovereignty.

Concerns over Claude Max’s $200/month pricing highlighted affordability disparities globally, impacting access for developers outside the US.

Interest in local quantized models like Qwen 3.5 and upcoming Deepseek was expressed, questioning if performance matches the hype.

Ethan Mollick (@emollick) questioned the alignment rationale for open weights models, covering both misuse prevention and broader human alignment for agents or AGI.

He clarified that governments lack access to superior AI models compared to the public—often lagging—and rely on the same cloud providers, with fine-tunes or SLMs not expanding general capabilities.

Mollick recommended Ada Palmer’s “hard social science fiction” like Terra Ignota and her early 2023 AI insights.