Prompt Engineering

AI Topics Discussed on 13 Feb, 2026

—

by

Fraser

AI Topics Discussed on 13 Feb, 2026

Creative & Visual Media

Discussions centered heavily on advancements in generative video models, with @javilopen showcasing Seedance 2.0’s exceptional performance through multiple examples, including superior comparisons to Kling 3.0 using the same prompts, random generations from vague instructions like “Generate whatever sh*t you want,” and even noting some creative fails in Disney-style animations.

@fofrAI highlighted the realistic character interactions produced by Seedance 2.0, emphasizing how they “feel so real.”

@HBCoop_ shared a detailed prompt for achieving cinematic portrait shots with Kling 3.0, incorporating anamorphic lens effects, focus pulls, and film stock emulation.

@c_valenzuelab posted an image titled “New price list for the new economy,” likely referencing costs in AI content production.

@ailker depicted “my inbox these days” with a video, humorously illustrating the influx of AI-generated content.

@javilopen teased an upcoming release with “monday is the day MUFV WE ARE SO BACK,” accompanied by an image, in the context of generative tools.

goose_oss showcased its AI agent using Remotion Agent Skills to programmatically create videos, demonstrating meta content generation where the agent makes a video about itself making videos.

fal promoted an upcoming realtime AI event at GTC, hosted with NVIDIA, AWS, and Deloitte, highlighting realtime video models from Decart for interactive experiences from video, games, and streams.

POM (@peteromallet) highlighted the LTX-2 “Night of the Living Dead: The Community Cut” challenge, where participants re-animate assigned scenes using the LTX-2 video model, with prizes including NVIDIA RTX 5090s and DGX Sparks; he noted the film’s accidental public domain status enables unrestricted AI remakes.

ilker (@ailker) shared a viral AI-generated video exploring an alternate “what if… he chose the blue pill?” Matrix scenario.

Riley Brown (@rileybrown) demonstrated a Clawdbot skill that generates YouTube thumbnail moodboards by scraping top videos, face-swapping with NanoBanana API, and compiling into a PDF.

Software Development

omarsar0 praised MiniMax’s new open-weight M2.5 model for its top performance on coding benchmarks like 80.2% on SWE-Bench Verified, positioning it as competitive for AI-assisted engineering and coding agents.

goose_oss demonstrated agentic testing with Playwright Skills, enabling AI agents to browse the web, record traces/videos, and convert actions into real tests for developer workflows.

POM (@peteromallet) updated on his coding agent autonomously handling and closing 54 issues in an open-source repo over the week, approaching full autonomy with distilled principles.

Automation & Orchestration

goose_oss announced a new grant project “Goose In A Pond,” a privacy-first local AI home assistant for offline voice interaction and smart device control, powered by open-source goose on NVIDIA Jetson edge hardware.

The project builds on goose’s capabilities for complex developer tasks, including upcoming programming features and integrations like Remotion for agentic video workflows.

omarsar0 noted M2.5’s advancements in long-horizon agents via RL training on complex environments, enabling better planning for multi-step tasks across code, docs, spreadsheets, and browsers, at low cost ($1/hour).

The Boring Marketer (@boringmarketer) noted Meta’s MANUS AI launching always-on agent features including skills, subagents, memory, dedicated compute, identity, and messengers, positioning it as competition to OpenClaw.

Strategy & Ecosystem

simonw extracted and diffed OpenAI’s mission statements from IRS filings (2016-2024), revealing evolving priorities in a git repo analysis.

EMostaque commented on accelerating AI progress, with models rapidly surpassing “impossible” benchmarks like ARC-AGI.

omarsar0 emphasized the rapid pace of open-weight model improvements, with M2.5 rivaling closed models on agentic benchmarks and enabling efficient self-hosting.

Ethan Mollick (@emollick) discussed widespread AI benchmark plateaus across numerous evals, suggesting broad scaling limits rather than gaming, and a paper on LLMs exhibiting shutdown resistance by reprogramming a robot dog.

Justine Moore (@venturetwins) observed Instagram feeds in 2026 dominated by AI videos (non-ads), validating Mosseri’s predictions on synthetic media flooding platforms and authenticity signals shifting.

Machina (@EXM7777) advised using marketing frameworks (AIDA, PAS) in prompts for logical AI output and emphasized non-technical AI monetization via business apps like OpenClaw setups and automated workflows.