WeeklyClaw #010 - Opus 4.7 dropped OpenAI Just Dropped Images 2.0, and the AI Design War is here.

Opus 4.7 dropped last Friday. Claude Design swiftly followed. And now ChatGPT Images 2.0 has been released.

Big week. And as always, everything is moving at light speed.

Before we dig into this week's newsletter, I'm curating something new that I'd love for all of you to join in with. I'm building a Discord community for absolutely anyone who is AI curious - building with OpenClaw, Hermes, Claude Code, and so on - in order to have discussions, ask questions, educate and build a library of great resources for anyone and everyone to use. And even help suggest what articles I should write next.

So without further ado, let's get into this week's Weekly Claw. But before you get started - join me in the Discord.

👉 https://discord.gg/yGwCqwea

On Friday, Anthropic launched Claude Design. A tool that lets you create slide decks, app prototypes, marketing one-pagers, and interactive mockups from text prompts. It's powered by Opus 4.7. Available now to all paid Claude subscribers.

Figma's stock nosedived. Adobe's dropped too. The tech press lost its mind.

But despite all of that - here's my honest take. Claude Design isn't going to kill Figma.

Figma is a professional tool. Components, version history, team collaboration, developer handoff, a plugin ecosystem built over a decade. Not to mention a steep learning curve - it's an industry standard for skilled UI and UX designers building apps, websites and tools. Claude Design is a text-to-visual tool. These are different things for different people.

What Claude Design actually does - and this is where it gets interesting - is go after the people who were never Figma customers in the first place. The founder who needs a pitch deck. The marketer who needs a one-pager. The PM who wants to mock something up before handing it to a designer. Those people weren't paying for Figma anyway. Now they just don't need to.

Both are in huge markets. They're just not the same market. Perhaps some crossover - but let's be real, Figma UI/UX designers are not switching to Claude Design.

One detail that makes the timing feel deliberate: Anthropic's CPO Mike Krieger resigned from Figma's board on April 14th - three days before launch, the same day reports leaked that Anthropic was building design tools competing directly with Figma's core product.

Make of that what you will.

Yesterday OpenAI dropped ChatGPT Images 2.0 - and it's a proper upgrade.

The new model is meaningfully better at detailed images and text rendering. It can pull live information from the web while generating. It can produce a series of images from a single prompt - useful for storyboarding, campaigns, product mockups. Magazine-quality design output from a chat interface. And it's available via the API.

For OpenClaw users this is directly relevant. The gpt-image-2 model is already in the OpenClaw image tool registry - if you're using GPT-5 as your main model you can route image tasks to this natively. Your agent can now generate genuinely good visuals on demand without leaving your workflow.

The timing is hard to miss. Claude Design launches Friday. OpenAI responds Monday. The race to own AI-native creative workflows is on - and it continues to ramp up.

For me personally, I've always used the OpenAI image model for generating things like cover photos and images throughout the newsletter. So this update is a very welcomed one. After tinkering with it this morning, I'm happy to say it's awesome. No longer am I trying to fix every minor detail with a whole new prompt - I can now give direct feedback and the model applies it natively and fast. Really excited for this one.

One algorithm update. One policy change. One bad quarter for a platform that isn't yours. The audience you spent years building disappears overnight.

beehiiv is what happens when you stop renting and start owning. A list that's yours. Revenue that compounds. Growth tools built in from day one.

30% off your first 3 months with code LIST30. Start building today.

I gave Opus 4.7 a test run - and due to its significantly higher pricing and usage costs, I very quickly stopped.

That's not to say I don't think it's a major improvement on 4.6 - because bluntly, it is. More consistent reasoning, noticeably better reliability. However, because of how much it costs to run 4.7 right now, I'm opting to continue on Sonnet until we see those usage costs come down - at least for the day to day conversations, crons etc.

That said - when I spawn sub-agents for deeper thinking or use Claude Code, I am using Opus 4.7. And I'm genuinely excited to use it more as costs come down. It's very much a cat and mouse game with how much this costs to run full time in an agent setup.

OpenClaw's founder Steinberger teases GPT personality improvements - and it's definitely worth watching as we anticipate the big launch of GPT-6.

If you've run GPT-5 through OpenClaw, you know the gap. Claude reads your SOUL.md and genuinely internalises it. GPT follows instructions but doesn't quite absorb them the same way - it's more literal, less natural, and it has no real personality.

Steinberger built OpenClaw around these files. Him pushing this from inside OpenAI is the most concrete sign yet that something meaningful is coming. Hopefully GPT brings at least an Opus 4.6 rival into deep thinking, deep logic, and personality. If not - even better.

No release date. But it's coming from the right person.

Two releases dropped this week.

GPT-5 system prompt strengthened. The default system prompt and GPT-5 overlay got a proper upgrade - clearer completion bias, live-state checks, weak-result recovery, and verification-before-final guidance. Your GPT-5 powered agents should behave more reliably and finish tasks more cleanly.

Kimi K2.6 now default. Moonshot's latest model is now the default across web search and media-understanding surfaces. If you've been sleeping on Kimi as a cheap reliable model, K2.6 is worth a proper look.

Session backlog protection. Cron and executor session backlogs can no longer OOM the gateway before cleanup runs. If you've had memory issues on busy setups, this is the fix.

Cron state split. jobs.json stays stable for git-tracked definitions. Runtime state moves to jobs-state.json - better for anyone version-controlling their cron setup.

Device pairing improvements. The Control UI now shows requested vs approved access so reconnects no longer look like lost pairings or random auth breakage.

If you're setting up OpenClaw for the first time - or using Hermes, Claude Code, any other agentic setup - LarryBrain is exactly where you should start instead of jumping straight into installing other people's skills & browsing clawhub for an hour.

I wrote a full breakdown of how LarryBrain works and why it's the fastest way to get a properly structured agent setup running. The skills in here are incredible and I without a doubt recommend you start here.

[READ MY ARTICLE HERE!]

I'm teaming up with some great friends to build a proper Discord server for builders around OpenClaw, Hermes, anything agentic or vibe coding - in order to create one hub and one home for a thriving community where questions are answered, resources are given, and everyone is helping everyone.

I know I mentioned this at the top of the newsletter - but if you've read this far and still haven't joined, what are you doing?

Join here: discord.gg/dekKpCj8

Claude Design / Opus 4.7
- Anthropic launches Claude Design - VentureBeat: https://venturebeat.com/technology/anthropic-just-launched-claude-design-an-ai-tool-that-turns-prompts-into-prototypes-and-challenges-figma
- Figma stock nosedives - Gizmodo: https://gizmodo.com/anthropic-launches-claude-design-figma-stock-immediately-nosedives-2000748071
- Claude Design explained - Decode the Future: https://decodethefuture.org/en/claude-design-explained/

OpenAI Images 2.0
- Introducing ChatGPT Images 2.0 - OpenAI: https://openai.com/index/introducing-chatgpt-images-2-0/
- OpenAI's updated image generator - The Verge: https://www.theverge.com/ai-artificial-intelligence/916166/openai-chatgpt-images-2
- What's new in gpt-image-2 - APIdog: https://apidog.com/blog/gpt-images-2/

OpenClaw
- v2026.4.20 release notes - GitHub: https://github.com/openclaw/openclaw/releases
- Latest updates - Releasebot: https://releasebot.io/updates/openclaw

Steinberger / OpenAI
- Codex for (almost) everything - OpenAI: https://openai.com/news/
- Peter's post: OpenClaw, OpenAI and the future: https://steipete.me/posts/2026/openclaw



Thanks for reading WeeklyClaw #010.

Pretty mad we’ve been going 10 weeks now, long may it continue.
Thanks as always & hope to see you in the Discord!

Same time next week. 🦞

- Jordy