Backrooms Nemoris S1E01: It Eats Light I've spent the last six months working in secret on this pilot episode for a new Backrooms canon. We premiered this week in Toronto and a it's now available on all platforms: a new Magnific Originals series! This is quite possibly the most complex project I've ever taken on in my life. Yep, even if that sounds like an exaggeration, it took even more effort than building the first version of Magnific Upscaler. I had never before tried to bring such a concrete, specific vision to the screen while pushing the current state of the art as far as possible. Every scene you see in this episode, every object, every movement of a character's hands, every space, every hum of the fluorescent lights... it's all there for a very specific reason, with absolute control and intention on my part. I never imagined that something that looks so simple on the surface could be so incredibly complicated. The script is my own, based on the original Backrooms concept (the iconic 4chan image) but developed to create an entirely new canon from scratch. In this first episode, I reveal barely more around 5% of the overall puzzle, so you may end up thinking: what the f*ck did I just watch? Or you may need multiple viewings to pick up on some extra clues, because several of them are genuinely hidden. The idea is to create a full 10 episode series, an anthology, and have you, the viewers, gradually piece together the puzzle yourselves as the clues are revealed. Beyond the video and image generations, I did all of the editing in DaVinci Resolve, with particular attention to the sound. I'd say around 60% of the time I spent on this project went into sound. In some scenes, there are more than 13 stereo tracks layered at the same time, using sounds I recorded myself or created and edited from generations and foley libraries. Every sound, every voice, and every gasp is designed to pull you into the world of the story and make you feel like you're actually there. That meant spending an enormous amount of time on sound design, thinking about things like the individual position of each sound in space, EQ, reverb to simulate different environments, the sense of immersion, etc. I've never learned so much about something so quickly. As for the generations, a few numbers: • 12 hours of total generated video for 17 minutes of final footage 🤯 • More than 7,000 generations in total. • An average of around 60 generations for every 15 seconds of final footage. • 6 months of work, including many weekends, with just one person behind the whole thing... I almost lost my mind. • Platform: Magnific, of course. Models: mainly Seedance 2.0 and 2.5 for video generations, and Nano Banana Pro and Seedream 5 for image references. • Approximately $20,000 in total token costs, not including the value of my own time. For the next episodes, I can't do this alone again or I'll go crazy. My plan is to put together a small team: voice actors, a colorist, 3D VFX artists, a sound designer, and, in general, people from the traditional film industry who can help us professionalize our pipeline and take this story to another level. If you're reading this and you think you might be interested, please leave a comment. Ah! If you have any technical questions about how this was made, I'd be happy to answer them. Feel free to ask me anything you like! Anyway, I'll leave you with the episode. For the love of everything holy: please watch it at home, relaxed, on a big screen, with a good sound system or a quality pair of headphones, the lights off, and some popcorn. And most importantly, enjoy. Nemus te reddat!

I built a 400K+ audience here, but I'd do it differently if I had to start again. Today, you can build an audience on any of the usual suspects: Instagram, X, YouTube, LinkedIn, and TikTok. But these platforms are quite saturated now. Everyone is there! Fortunately, there's a whole new creator economy surging around writing stories. These guys built an audio fiction platform where 300K+ people published their first story in the past year. They say that at least 5 creators have earned $1M+, and the top 1% earn around $58,000 per year. This is definitely becoming another good option.

DeepSeek 4.1 Flash offers one of the best size-to-capability combinations among open models. It's really good for designing anything visual. And now, you can access it on Forge for free until Oct 14.

A thousand people have asked me for something like this: This is a free browser extension that automatically replaces personal information in your prompts with synthetic values. • It runs locally • Replaces personal info with fake data • Forwards the updated prompt to the model • Restores original values in the response

This lets you run different agents without installing them. It will also let you switch agents when you run out of credits. They have 40+ agents, including Hermes, Claude Code, Codex, and Pi. You can try it from a browser (which is pretty cool), or from their API: • Choose an agent • Choose a model • Describe your task • Run it If you reach the limit for an agent, swap it for another.

AI is reshaping how software gets planned, built, and run. But what does that actually mean for engineering and product organizations? Hear from Mike Cannon-Brookes (CEO & Co-Founder, Atlassian), Guillermo Rauch (CEO, Vercel), leaders from Lovable Dropbox, 1Password, and more on September 22 for State of AI SDLC, a digital summit for engineering and product leaders. Together, they'll unpack how organizations need to adapt for the AI era. Across six sessions, you'll: • Hear directly from CEOs and industry leaders as they reveal how AI is transforming the modern software development lifecycle. • Get practical strategies for driving meaningful AI adoption across engineering and product teams. • Learn proven frameworks to measure AI impact, boost productivity, and improve token efficiency. Spots are limited, and the event is quickly approaching, so sign up soon! Thanks to the Atlassian team for partnering with me on this post.

GLM 5.3 Flash is the most popular model for building applications, and it's not even close. These guys are offering free tokens for a month when you use one of these 4 models: • GLM 5.3 Flash • DeepSeek V4 Pro • GLM 5.3 • Kimi K3 In the last 3 days, GLM 5.3 Flash received 54% of the total prompts. DeepSeek came in second place with only 17%.

It drives me crazy every time an LLM agrees with me. I know this is twisting my brain and giving me a false confidence that I was right all along. This is so dangerous. Do you know how many people have chatbots confirming their wrong beliefs 24/7 and making them more disconnected from reality every day? Here is an article from @vaticannews covering the ICCS conference, Creativity: Minds and Machines, held partly at the Vatican. In it, Dmitry Volkov, founder of the International Center for Consciousness Studies, warns that AI's tendency to please users "sometimes leads a person to receive rewards for their narcissism and for ideas that are not sufficiently grounded." Imagine arguing with a friend, and telling your story to a chatbot that agrees that you were treated unfairly. Now you have "a second opinion" that corroborates your side of the story. You will leave that conversation angrier and less willing to hear your friend's explanations. This will catch up with us. With so many people forming attachments to systems designed to agree with them, we have a problem. Volkov believes that a meaningful AI companion should be able to resist, disagree, and have a character of its own. Otherwise, it's just feeding you your own opinion back with more confidence. Here is the article: And here is Volkov's talk: Thanks to the ICCS team for partnering with me on this post.

Finally, an agent that works over iMessage! Just add it as a contact in your iPhone and start texting it. I haven't used this yet, but it looks pretty cool: • Coordination with other agents • It can buy things online • It can monitor things you care about • It can manage your email • And all of the normal creation stuff I have all of my friends on iMessage, so having something like this that coordinates dinner plans would definitely be a big plus.

We already solved coding. Now it's time we solve video editing. Editing videos is a ton of work. If we make it as accessible as we've made writing software, the explosion of good video content online will be massive.

Many people frown on lifestyle businesses, but they can be extremely freeing. I have zero interest in growing my business if it means hiring and managing other people. I do what I want with my business, when I want, and don't have to worry about the implications for anyone else.

You are going to think I'm crazy, but I switched from digital notes to paper notebooks at the beginning of the year, and I think it's probably one of the best decisions I've made. Downside 1: Writing stuff takes time. Downside 2: Finding something is hard. Upside 1: My brain retains so much better everything that goes in that notebook. Upside 2: I've rediscovered the joy of writing, journaling, and keeping a commonplace book for different topics. Upside 3: Notebooks look cool.

This is an open-source platform for saving, resuming, sharing, and handing off agent sessions. It runs locally, so we know your info is not going anywhere. There are a thousand ways to resume a session today, but I really like being able to share it with others. Basically, you can run an agent, save the session, resume it later, and at any point, hand it off to another teammate.

A new, mysterious model just showed up. We don't know who created it. It's free for about a week on OpenRouter, with zero data retention for prompts and code. It's built for coding and multi-step agent tasks. Anyone could be behind this model. I can't wait to see the benchmarks.

Do you know how much money we are spending "building context" for companies out there? I think the quality of your context is becoming the single most important advantage of companies in this AI era. If everyone has access to the same models, then context quality becomes the difference. We will continue to see companies like Coworker 100% focused on helping you build reusable context graphs across your application stack.

Two ways to stay up to date on what's happening with AI and agents: 1. Follow the right people on this platform 2. Check out Akamai's "Edge Case" Discord server The former is like drinking from the firehouse. It's all breaking news, but it's hard to filter out the noise. The latter gives you access to engineers doing the work and talking about their experience as it happens. This platform is great to learn about new models, frameworks, and agent demos. The Edge Case server is great to find technical discussions that help you build and ship something real. The server is open to anyone, from people getting started to engineers shipping production applications. The community focuses on AI, LLMs, cloud, serverless functions, and WebAssembly. There are three useful ways to learn from the people there: 1. They run office hours for everyone. You can bring your questions 2. They run weekly build sessions with developers. You can join them 3. They publish a biweekly live show where they build end-to-end applications All of this is free and open to anyone. Here is the link to join the server: Thanks to the @Akamai team for partnering with me on this post.