DevArcher (@katchimeradev) header
DevArcher (@katchimeradev) profile picture

Building SynArc | Prompt → world → everything else. AI-native worlds with memory, continuity, stories, systems & cinematics.

London, England
Joined May 2023
Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)Photo by DevArcher (@katchimeradev)
New challenge - I will build a midcore mobile game at the depth of some of the biggest merge and 4x games out there: the gameplay and collection depth , the live ops tech and event cadence , the content, you name it. This stuff takes big teams. This is my profession , I worked on a billion $ mobile franchise and know how it works With AI, it’s possible at a fraction of the manpower and I will share that journey starting on my own
3
2
41
Ideally when this is finished or I get publisher support I would look to redo art and narrative text with actual artists for player trust , but when we get there The amount of iteration is huge though nothing about this is one shot , ai simply can’t get these things right without the human in the loop iteration , taste and direction
12
It’s still rough around some edges but I wi sure the progress here :)
2
Here’s the opening sequence so far All the art is AI so far The first time user experience tech , the data driven nature of everything , the live ops backend support to bundle over the air builds , it’s all there ! Let’s see how this goes It’s a life companion merge 2 x 4x games- Katchi is the active name for now :)
28
Haven’t posted in so long but wow how the space has moved I’m going back to my roots as a game dev and b2c builder and sharing the journey here
23
Found a workflow that lets Claude Code agent generate any real production asset you want: images, video, transparent PNGs, directly from your own backend. No more manually wrangling graphics. This was with Fable, before it was sadly suspended but I think it can be done with Codex gpt 5.5 and opus 4.8 Here’s the full setup 🧵
5
103
Fable generated all assets it needed automatically without even prompting it directly , just did whatever it needed as part of the feature work , even setup birefnet by itself , which was amazing for the iOS app I was building -which involved creating creature card graphics with transparent backgrounds and more! I don’t know how well opus 4.8 or got 5.5 would handle that , but at the least we can prompt them to do things more directly if need be using this setup
36
Tell Claude to build edge functions that call fal.ai using that secret. These become your asset-generation endpoints: images, video, whatever fal supports. Then just prompt. “Generate a hero image.” “Cut out this background.” Claude hits the edge functions directly and the finished assets come back ready to drop into your app. The transparent-background part is underrated, it runs BiRefNet to cleanly cut out subjects. Huge for app UIs where you need proper PNGs, not boxy stock images.
42
Create a fal.ai account and grab an API key. Add it to Supabase as a secret, not hardcoded. Keeps the key server-side and out of your client code. Run npx supabase login so Claude has authenticated access to your project. Now it can actually deploy edge functions, not just suggest them.
42
Spin up a Supabase project to act as your “command center.” This is where your edge functions, secrets, and asset pipeline will all live. Hit Connect, copy the prompt into Claude code, and install the Supabase skill. Now Claude knows how to work with your backend natively instead of guessing.
41
synarc.dev Here’s a preview of the SynArc landing page! If you’re a world-builder, visual storyteller, or filmmaker, sign up for early access !
32
DevArcher retweeted
The quality of films using AI gets better everyday. For genres like scifi and fantasy, which were usually expensive (unlike horror, which has always been cheap), we are about to see a huge renaissance.
5
6
48
2,310
It’s because most people using these novel tools are here for the tech, the novelty, or trend chasing to make a quick buck. There’s a reason so many crypto/NFT folk jumped into AI space, and there’s a reason AI UGC has exploded. But this will change as seasoned writers and artists start to finally adopt the tools. It’s already starting to change, I’ve seen some bangers, but I’ve yet to see quality scriptwriting for a lot of it. Also it tends to be that the most talented pro writers and artists are anti AI, given how much it infringes their work in their eyes. The result is many AI filmmakers are not actually traditional artists to begin with, but more tech-forward-minded people. We’ve yet to see the best with this new medium, especially as the traditional talent finally adopts it as the stigma dies out!
I'm probably being very harsh here, but I think most AI filmmakers don't think much about existential questions about life, society, sociology, and economics. As a result, they don't have really interesting things to say. And thu’s not very interesting films. It’s notable for being cheap or fast but mostly gimicky. If you don't have an interesting thing to say about life in society (or experience of it), you're not going to make interesting stories. You can knock Hollywood all day for being woke and DEI but the truth is that’s a legit perspective on society. It’s a critique of power and identity. That’s why I love Taylor Sheridan. He’s not woke or DEI but he has a real perspective that he clearly thinks a lot about. Those stories resonate.
62
DevArcher retweeted
the next massive consumer ai opportunity is making personal agents feel as intuitive as an iphone. this is deeply important because this is the new software layer for everyday life. most ppl do not want to configure workflows, manage prompts, route models, or think about agents at all. they want software that just works & the winning products will hide almost all of the complexity with taste incl. context, memory, & orchestration. e.g. there’ll be baseline personal agents that come alive out of the box which are already understanding your context, patterns, relationships, preferences, apps, devices, routines, etc. then there’ll be ephemeral agents that spawn dynamically from intent, ambient capture, conversation, location, screenshots, email, calendar, camera roll, whatever. this is the software that assembles itself around the moment just like weather updates based on your location but way more in depth. today even the most state of the art agent products feel like giving normal people shell access to a distributed system. apple won by turning computers from something you operated into something you experienced. personal agents require the same transition. whoever solves this becomes the ambient operating system for human life. small category btw.
140
106
46
1,496
274,836
SynArc on the journey to make ‘AI Pixar’ possible for long form feature length new animatic continuity tests It’s becoming clear we need a scene graph to improve long form continuity and SynArc will get it It would solve the hatch door artifact Prompt below 👇
2
36
Something we noticed helped with character performances is using the ‘valence / arousal’ scale
22
This prompt was autogenerate by SynArc So was the shot assembly from script, the shot graph was auto generated and hooked all references from the story world graph - the automated prompt generation is still open to improvement Prompt : [REFERENCE LEGEND] @Image1: primary opening keyframe reference. @Image2: Tansy Mott default reference; entity identity, wardrobe, variant, or prop continuity reference. [DIRECTED CONTROLS] Camera: wide lateral follow, slight high angle, subtle handheld urgency. Subject motion: sprint left to right, brief skid, grab and wrench hatch wheel, slip through opening. Focus: hold on Tansy and hatch. Framing: wide action frame, keep ledge and hatch visible throughout. Visibility: Tansy fully visible until entering hatch; no other characters visible. Performance: frightened, desperate, low-centered run; straining pull; no hesitation after hatch gives. Voice: hard breathing only, no dialogue. Motion: high. [SHOT] 00:00-00:05: She runs the ledge, slips slightly, grabs the hatch wheel, and hauls the hatch open. Lighting: Brass chamber glow spills across dusted metal, with deeper darkness visible. [PERFORMANCE / VOICE] - Tansy Mott: role: A gifted dormouse mechanic whose habit of fixing everything alone is tested; identity: Small dormouse mechanic with cinnamon-brown fur, a cream muzzle, oversized amber eyes; traits: young adult, very small stature, cinnamon-brown fur, cream muzzle, oversized amber eyes; voice: Bright warm alto with nimble diction, quick problem-solving energy, and a tremor of earnestness when feelings catch up to her Accent: soft harbor-town accent. [AUDIO] No music, score, audio bed, room tone, crowd wash, or background ambience. Use only scripted dialogue and direct diegetic sound effects caused by visible or explicitly offscreen shot action. Do not render production-board artifacts: no arrows, labels, captions, subtitles, guide boxes, panel borders, grid gutters, map diagrams, UI, logos, watermarks, or handwritten notes. Preserve attached refs and keyframe composition.
154