@AddictiveColors
Account based inUnited Kingdom
About this account
- Account based in
- United Kingdom
- Connected via
- United Kingdom App Store
Account-level information from X, not a live location or the device used for a specific post.
Developers of the Pocket Boy app for iPad. Now available on the Apple App Store. Please follow our blog too: https://nitter.cf/t.co/oF6GOi8kvD
Joined July 2011
- Tweets30.4K
- Following1.1K
- Followers498
- Likes46.8K
Addictive Colors retweeted
holy sh*t. Official Anthropic tip for Claude Code: stop burning Opus 5.5 context on tasks Haiku 5.5 can swarm, while Sonnet 5.5 builds
put the architect on call with custom subagents
CLAUDE_CODE_SUBAGENT_MODEL=haiku claude --model sonnet
Sonnet 5.5 drives the primary session on high effort, writing diffs and running test suites
Haiku 5.5 subagents swarm the repo in parallel at 340 tok/s, handling file discovery and spec docs lookup
Opus 5.5 stays on call in ~/.claude/agents/ as the code reviewer, only stepping in at three critical moments:
→ before a plan locks: does the implementation plan miss auth invariants or schema contracts?
→ when a test breaks twice: are we fixing the root cause or falling into a recursive rabbit hole?
→ before calling done: did the full diff introduce hidden regressions or break pre-flight rules?
Opus 5.5 reviews. Sonnet 5.5 builds. Haiku 5.5 swarms
Claude Code subagent architecture runs the identical pattern through isolated contexts: high-volume operations (reading 30 files, grepping AST, running test suites) execute in separate worktrees, so 400k+ tokens stay in subagents while the primary session keeps a pristine 14k context
Plan on high. Delegate on medium. Keep Opus on call.
- the full subagent swarm setup
> Opus 5.5 on call in ~/.claude/agents/code-reviewer.md catches deep architectural traps
> Sonnet 5.5 lead drives edits, writes core logic, and executes test harnesses
> Haiku 5.5 subagents swarm AST parsing, grep scans, and API docs in parallel at 340 tok/s
> isolation: worktree guarantees test runs and scratch edits never pollute your git tree
> isolated context windows shield the primary session from 400k+ tokens of verbose output
paste the setup and this prompt into Claude Code below:
"Configure my Claude Code workspace for hierarchical subagent orchestration:
1. Audit ~/.claude/settings.json and project config for model roles fitting lead, subagents, and reviewer:
> Pin main session model: sonnet with effortLevel: high for primary execution
> Set CLAUDE_CODE_SUBAGENT_MODEL to haiku in settings.json with CLAUDE_CODE_SUBAGENT_MODEL_FORCE to 1
> Pin custom code-reviewer model: opus with effort: high in ~/.claude/agents/code-reviewer.md
2. Create specialized subagent files in ~/.claude/agents/:
> code-reviewer.md: model: opus, tools: Read, Grep, Glob, read-only review of plans and diffs
> explorer.md: model: haiku, tools: Read, Grep, Glob, fast codebase search and AST discovery
> researcher.md: model: haiku, tools: Read, WebFetch, external documentation and specification lookup
3. Configure automatic delegation checkpoints in ~/.claude/CLAUDE.md:
> Invoke @ code-reviewer before finalizing multi-file architecture plans
> Automatically summon @ code-reviewer when the same test or compiler error fails twice
> Enforce @ code-reviewer diff contract audit before declaring tasks complete or staging git commits
4. Constrain subagent scopes:
> Enable isolation: worktree for subagents making edits or running test harnesses
> Return structured summaries only without flooding the main conversation context
Show every configuration diff first. Do not apply edits until confirmed"
↳ code.claude.com/docs/en/sub-…
Addictive Colors retweeted
New for Rex testers. 👁️ Activity Mode shows you active and idling processes and agents for quick context switching, and some extra metadata.
We're working our way slowly through the waitlist. Our priority is building quality software, so we're admitting testers in batches of 50 or 100 at the moment and addressing any feedback with the group before adding the next wave.
Addictive Colors retweeted
I have discovered $50 amazon photo frames to be the PERFECT home display screen
They can be rooted, updated to run a modern browser, have WiFi and can be powered from USB
1280×800 resolution with a nice touch screen
I detailed how to hack these yourself in my latest video
wesbos.com/hacking-smart-pho…
Addictive Colors retweeted
We’ve added Steam Frame support - available now in the latest #NoMansSky experimental build!
Eye-tracked foveated rendering is supported and should also work on other compatible headsets, including Meta Quest Pro and Pimax Crystal.
steamcommunity.com/app/27585…
Addictive Colors retweeted
Read the spec, we have support for hierarchical process trees: superlogical.com/rex/docs/bu…
Addictive Colors retweeted
I've published a new, general-purpose terminal specification that gives any program a way to tell the terminal what it's doing: idle, working, waiting, finished, or failed, and why. It is easy to implement on both sides. I've written more about why: mitchellh.com/writing/progra…
The problem this solves is generic and applies in compelling ways to all sorts of use cases (e.g. package managers, build tools), but it really rears a hideous head when it comes to AI and the "agentic inbox" problem.
I found over 250 different agent orchestrators (with various goals) that all individually implement a heuristic based approach to detecting whether things like Claude Code are working, blocked, done, etc.
If you look at @herdrdev's commit history, you can see around 10 compatibility changes to detect Claude Code within just the past 3 months. (This isn't a dig at Herdr! They're doing great work with what they have, and everyone else is doing the same thing).
Heuristics are not the way. And over-indexing on agentic use cases isn't the way, because this problem is really interesting for other tools too: Homebrew, Terraform, Cargo, etc. etc.
Proprietary protocols and out of band APIs are not the way. They have an O(N) integration problem and require extra work to work over SSH or within local VMs/containers.
The program status specification (OSC 7501) solves this in a reliable, deterministic way that is general-purpose and can be easily adopted by the entire industry.
Read more: mitchellh.com/writing/progra…
Addictive Colors retweeted
Haiku 5.5 is now available in the Claude Platform and Claude Code. On average, it costs around 75% less to run than Haiku 4.5.
It pairs well with Opus 5.5 or Sonnet 5.5 as a subagent. Use it for high-volume, cost-sensitive tasks like summaries, compactions, or database queries.
Addictive Colors retweeted
Yes! And TMOG is fully native on THREE different platforms (Windows, Linux, and Mac), with no Electron or other browser shell.
It means the Windows version uses Direct2D/DirectComposition. The Mac version uses UIKit/Swift, and the Linux version is C++/Qt.
Electron makes sense as a labor saving device, but labor is now almost free, so everything should be native!
tmog.org
Addictive Colors retweeted
>Berates her comrade for potentially alerting the enemy
>throws a massive flare down a cave
Addictive Colors retweeted
Ep. 30 is live! 🎙️ Big update! 🎙️ I'm finally ditching the AI voice clone and using my REAL voice. Why? My friend John Linneman gave me the perfect push: "You can speak English, so use your own voice even if it's not perfect." So here I am—imperfect but 100% real! 🗣️✨ I'm joined by ex-Capcom developers Mr. Oi and Mr. Kimoto to discuss the intense feedback culture that drives Capcom's amazing in-game development. Plus, I talk about my first time eating raw whale sashimi in Obihiro! 🐋
#GameDev #Podcast #Capcom #Ochify #GameIndustry
podcasts.apple.com/jp/podcas…
Addictive Colors retweeted
nitter.cf/mizuki_izuna/status/21…
CAPCOM’s next-gen RE ENGINE work is very interesting.
They’re basically attacking several pain points at once: C++ complexity, long build/link times, memory corruption and leaks, awkward reflection, duplicated C++/C# type systems, expensive marshaling, IL2CPP debugging, GC pressure, and memory fragmentation.
The answer seems to be RE:Runtime + RE:C++.
RE:C++ is built around C++23, but with C#-like semantics layered on top: automatic memory management, generics, interfaces, boxing, exceptions, and access to the .NET standard library.
The really interesting part is the workflow.
CAPCOM is using C# as the frontend language for parts of the runtime. Code is written in C#, compiled into a normal .NET assembly, then RE:Build converts/decompiles that into RE:C++ and builds it into native libraries/executables.
So the rough pipeline is:
C# → .NET DLL → RE:C++ → native C++ binary
That means developers can work with C# syntax, tooling, libraries, reflection, attributes and unit testing, while the final game/runtime code still ends up native.
They’re also doing their own memory-management work underneath this. Instances are managed in 64 KB pages, with page reference counting plus cache and GC optimizations. The goal seems to be reducing the cost of managing huge numbers of objects while improving locality and avoiding some of the fragmentation problems they’ve had before.
This also makes CAPCOM’s REDox project much more interesting.
REDox is their open-source .NET serialization/data library, with JSON, CBOR, MessagePack-style structured data handling, lazy decoding, streaming, compact token storage, etc.
At first glance it just looks like a standalone .NET library.
In context, it looks much more like the kind of infrastructure they can write and test normally in C#, then feed into this RE:Runtime / RE:C++ pipeline and ship as native engine code.
#RE2026CAPCOM #CAPCOM
Addictive Colors retweeted
Distributed systems are so cool, but also one of the biggest challenges to get *just* right.
Consensus protocols. Connection multiplexing. Caching. Failover. So many factors to consider that require careful design choices.
Been so cool to see Neki come to fruition over the past year+, and I'm blessed to work with some of the best distsys engineers in the world.
A snippet from my talk at Neki night a few weeks ago.
Addictive Colors retweeted
Capcom’s REX project is a gradual evolution of RE ENGINE, not a replacement. The goal is to modernize core systems while keeping existing projects and workflows intact.
RE:Dox: standardizes game data for faster loading, easier editing, and better interoperability.
RE:UI: rethinks tool UI architecture to reduce freezes, GC pressure, memory use, and unnecessary updates.
RE:Log: compact ID-based binary logging, centralized collection, localization support, telemetry, and integration with RPC.
RE:Flows: visual gameplay authoring that can be converted into a standard programming-language representation for easier collaboration and maintenance.
RE:Runtime: new runtime and memory management designed for very large numbers of game objects.
RE:C++: an intermediate or target language intended to translate developer-facing code into efficient runtime code.
AI integration: Capcom wants RE ENGINE to become an “AI-generation game engine,” where AI can understand code, generate implementations, run tests, and find problems automatically.
The interesting part is that Capcom is doing this incrementally. Instead of building “RE ENGINE 2,” they are replacing individual systems as needed while continuing to ship games on the same engine.
カプコンのゲーム開発を支える「RE ENGINE」はどう進化するのか…新プロジェクト「REX」が目指す未来
gamebiz.jp/news/433531
＝＝
Addictive Colors retweeted
bro aint wrong
We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:
Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:
Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:
Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:
Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!
In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
Addictive Colors retweeted
We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:
Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:
Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:
Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:
Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!
In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
Addictive Colors retweeted
Once upon a time we had quality standards on pixels, like in the flight sim industry, accuracy in sub-pixel motion, energy conservation. Then ML shows up and now temporal flicker and trailing edge ghosting are somehow ok?
Addictive Colors retweeted
Sony’s PSSR2 SIGGRAPH talk was great. The old version was pure RGB neural net. AI learned interpolation and filtering and color space conversions, etc. Now they just do kernel prediction by AI. How to blend history. Rest of the technique is traditional computer graphics algos. It’s faster and has much higher quality. PSSR2 proves that domain knowledge still matters. They only offloaded the part that was tricky to generalize to AI. We have been hacking/tweaking those TAA parameters for decades. Same thing with neural materials. Don’t let AI to learn signal processing (fourier transforms, etc), use fast human written code for those. AI doesn’t have to be brains-off lazy. Effort still pays off, especially in real-time graphics.
Addictive Colors retweeted
Brilliant engineer. Hire him 👇
A few days ago, my journey at Goodnotes came to an end.
It's been almost 6 years working alongside the most talented people I've worked with in my entire career, and we even won a Apple iPad App of the Year Award together in 2022!
I'd like to reflect a bit on some of the most important projects that I've worked on in the last few years:
- I started working on core features of the app, like the initial version of Comments or the in-app Notification Center.
- Soon after that, I was involved in the transition of the app from being paid upfront to having a freemium model, unlocked by in-app purchases. This helped us grow our user base considerably. Later on, I worked on a similar transition to a subscription model, enabling recurring revenue.
- A Hackweek project of mine resulted in the creation of the Pop-up Shop (we managed to put this into production in 2-3 weeks back when there was no agentic coding) and later on, the Marketplace, both selling digital stationery to be used within the app. These projects helped us gain new sources of income.
- I led the design of a DRM system to protect the IP of the content we were selling through the Marketplace (probably one of the most challenging projects I've ever faced).
- In the last couple of years, I moved to the AI team, where I kickstarted the Freeform Agent, an AI agent that's able to edit freeform documents, and numerous other features in this domain. I even moved out of the iOS ecosystem and led the projects to create ChatGPT and Claude apps for Goodnotes.
- On a more general aspect, I've tried to create a knowledge sharing culture by hosting the Engineering Brownbag Sessions, where every other week, our engineers where presenting their learnings in projects both inside and outside the company. This even got me the "Share Your Passion" Award inside the company!
- And finally, on the fun side, I created the tradition of playing Blood on the Clocktower in our company offsites. It was truly fun and I hope they keep doing this on my absence.
I will be taking some time to rest now, but I'm looking forward to my next adventure!
Addictive Colors retweeted
Coinbase is going native too. Judging by how many teams reached out after our native announcement, expect more of these soon.
