
Wednesday, Oct 23
Presented byArcjetHappy Hump Day! In today's Leaderboard: Runway's newest generative video tool could make animation studios very happy, an app that makes it easier to finally sell all the things gathering dust, and a new AI code review companion. Let's dive in
Make your own Toy Story
Runway Act-One: Generate expressive character performances with AI.
Runway Act-One is the AI company’s latest addition to its generative video suite, enabling expressive character performances across different genres by using both video and audio as inputs. Watching the demo, it’s definitely impressive—if I were Pixar, I’d be pretty excited. Still, with how saturated the market is, I don’t feel the same sense of wonder as I used to. Maybe I’m just getting a little desensitized to AI. Anyone else?
Bank account go brrrr
Hero: Lets you identify, price and list items for resale simply by taking videos of them.
I love the concept behind Hero, which was made by former Product Hunt engineer Rahul. When I tried the app, the first few things that popped out didn't seem to accurately reflect what's on the videos — e.g. Nikes in the video for $30 but the listed price shows $58, or an LG computer monitor listed as an Algae monitor. This isn't a dealbreaker, though, since I'm willing to tweak a few things if it can take care of the burdensome parts (image/video/title/description and estimated range of prices).
Too much noise...for now
Trag: An AI-powered code review companion.
In theory, I like the idea of enforcing linting and best practice rules that are too complicated for linters to enforce. In practice, Trag didn’t work as well as I’d hoped. I tried setting up a few basic rules that ended up in a bunch of false positives and added noise to our PRs. But the rules I used were pretty generic and I think that's why the results were underwhelming. There's probably a sweet spot here: rules that are too complex for a linter but not as complicated as architecture or design.
Your agents stopped answering and started doing
Your agents call tools, write to databases and move money, so a wrong answer is now a wrong action, which is a considerably worse meeting. Arcjet runs policy before every LLM call, tool call, query and API request: prompt injection, sensitive data, bots, rate limits, spend quotas, and your own rules in Rego on top. It reads the OpenTelemetry you already emit, so switching it on is one environment variable, no code changes and nothing to deploy. Each check also reads the earlier steps in the same workflow, so an agent's tenth action gets judged against the first nine and not on its own. JavaScript, Python and Go SDKs all at 1.0, agent framework guards included, free tier to start on.









