Arthur Soares

@GregMolnar Try aerospace. github.com/nikitabobko/Ae…

Not gonna lie, MacOS is such a downgrade from Omarchy. I will try to get a tiling setup, but I don't have high hopes.

@petergyang I spent the last 48 hours in this deep process of trying to amend it. It seems that the agent and soul migration from Claude to ChatGPT got lost in translation. The two models interpret things completely differently

I know OpenClaw isn't part of OpenAI but this feels like a mini-crisis for OpenAI if the GPT integration doesn't improve soon. The bar is GPT needs to be just as good if not better for OpenClaw as Opus.

@steipete @openclaw I have both settings enabled. Unfortunately, it is still a struggle. I have to constantly ask for checkpoints and nudges so the agent continues executing the plan. :/ I’ve already completly rewritten my agents/souls for more “gpt-like” language to no success. :(

Two experiments in the next @openclaw to address some "GPT is lazy" issues: 1) Strict mode: agents.defaults.embeddedPi.executionContract = "strict-agentic" This tells GPT-5.x to keep working: read more code, call tools, make changes, or return a real blocker instead of stopping at “here’s the plan.” docs.openclaw.ai/providers/open…

@ameliom612 I use PinchTab

What is the best browser to use with openclaw? I have multiple agents and want each of them to have access to a browser. I've been giving each their own chrome profile but its a headache when it comes to having to sign into the same things on each one. When they share one profile they step over eachother and close out other tabs like they are fighting over it. #openclaw

@metageg @Pickering @openclaw Any additional findings?

@artsoares @Pickering @openclaw super interesting, rhanks for doing the experiments!

OpenClaw 2026.4.10 🦞 🧠 Active Memory plugin 🎙️ local MLX Talk mode 🤖 Codex app-server harness plugin 🧾 Teams pins/reactions/read actions 🛡️ SSRF hardening + launchd fixes stability, but with attitude🦞 github.com/openclaw/openc…

@rileybrown This is the way.

Personally I don’t see the point of Cowork. I don’t even see the point of chatgpt or Claude anymore. I either want to talk to a claw (agent with all my skills running on a persistent computer with NO guardrails in iMessage or telegram) or i want to use Claude Code or Codex app for coding.

@Saboo_Shubham_ @jeffscottward For which cases are you splitting tasks/domains between Hermes and Openclaw?

@jeffscottward I'm already using both. OpenClaw actually talks to my Hermes agents for things that it is good at.

OpenClaw with GPT-5.4 feels noticeably better with the new update. GPT-5.4 (thinking=high + fastmode=true) seems like the sweet spot. Better judgment, less back-and-forth, still fast enough for daily use.

@Shpigford I had a lengthy fight with it this week, it seems that openclaw is model-agnostic, but the different models really think of the set of instructions in a completely different way. It felt like suddenly working with a different team. arthur.earth/it-feels-like-…

i turned off my 🦞 the past week of trying to force it to work with other models was a massive drain and i was able to replicate the vast majority of impactful functionality with a mix of claude cowork + code.

@Pickering @openclaw I wrote a lengthy write-up on a similar experience while migrating from Claude models to GPT. arthur.earth/it-feels-like-…

I’m running mine on ChatGPT-5.4 Oauth. For some reason my little molty’s personality has fallen off a ciff in the last week. I’ve messed with the settings from various tweet recs and perhaps one of those might have been the original reason. Is there a recommended setting that might bring his rizz back? Or something that perhaps Codex could debug? Thanks!

Writeup + benchmark: arthur.earth/it-feels-like-…

Migrated my 14-agent household assistant from Claude to GPT-5.4 after Anthropic cut off third-party harnesses on April 4th. "Model-agnostic framework" ≠ "model-agnostic instructions." Took a full day of rewrites.

Framer started as a tiny CLI tool for adding borders and captions to photos, inspired by a box of old family prints. A year later, it’s become a native SwiftUI app for macOS and iOS, with layers for borders, captions, dithering, LUTs, shaders, and more. arthur.earth/framer/

Remember reading CD booklets while listening? That ritual died with streaming. Built a now-playing display that brings it back: beautiful album art on any tablet, plus AI that surfaces facts like digital liner notes—recording stories, lyric origins, studio gossip.

Beginnings with Photography arthur.earth/beginnings-wit…

We're in for a ride arthur.earth/were-in-for-a-…

For the confused, it's actually super easy: - GPT 4.5 is the new Claude 3.6 (aka 3.5) - Claude 3.7 is the new o3-mini-high - Claude Code is the new Cursor - Grok is the new Perplexity - o1 pro is the 'smartest', except for o3, which backs Deep Research Obviously. Keep up.

For all Leica photographers out there shooting with 3rd-party lenses, I've made a script that helps automating the process of fixing EXIF data to the correct Lens, so you can get proper Metadata and Lens Correction in Lightroom. gist.github.com/arthursoares/6…

.@diveshnaidoo shared a new mind-blowing demo showcasing the capabilities of @Simulon, an upcoming app that allows you to easily add 3D models to real-life footage. More demos and info: 80.lv/articles/a-coo… #simulon #vfx #visualeffects #3d #3dart #3dsoftware #gamedev

@duborges You should check JourneyBase.io

Tired of Discord? This app lets you generate Midjourney images much faster. Also has a super handy Prompt Builder.