Full Story w/ Open Brain Agent Memory: https://natesnewsletter.substack.com/p/openclaw-agent-runtime-model-swapping?r=1z4sm5&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true ___________________ What's really happening inside OpenClaw when everyone is arguing about the model layer but missing that the runtime itself changed shape in April? The common story is about Anthropic versus OpenAI and subscription policies — but the reality is that OpenClaw crossed into serious work mode, and once you can swap brains through a durable work layer, memory becomes the strategic layer that matters most. In this video, I share the inside scoop on what April's releases actually mean for builders: • Why OpenClaw is becoming a runtime abstraction for serious agentic work, not just a chatbot wrapper • How Anthropic's subscription changes and OpenAI's Codex access create opposite architecture assumptions • What makes a durable workflow survive model churn, pricing changes, and better local models • Where OpenBrain for OpenClaw fits and why memory can't live inside any one brain Leaders treating model choice as a permanent architectural decision are missing that the practical unlock is designing workflows that outlive a provider policy. Chapters 00:00 OpenClaw grew up in April 02:30 From viral demo to serious runtime 05:00 The boring stuff that makes work possible 07:30 Task flow, memory, and channel maturity 10:00 Anthropic's April move was deeply unpopular 12:30 OpenAI's opposite posture with Codex 15:00 Gemma 4 and the local model branch 17:30 Which model should handle this step 20:00 Durable workflows that survive the session 22:30 Memory can't live inside one brain 24:30 OpenBrain recipes for OpenClaw 25:30 Build the runtime so the model can change Subscribe for daily AI strategy and news. For deeper playbooks and analysis: https://natesnewsletter.substack.com/ Listen to this video as a podcast. - Spotify: https://open.spotify.com/show/0gkFdjd1wptEKJKLu9LbZ4 - Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-news-strategy-daily-with-nate-b-jones/id1877109372
ADVERTISEMENT
5 minutes of insight hidden in 32 minutes of content.
Thanks for the review Nate. Do you think the future is really in swappable brains for workflows? I feel like this is similar to one hardware running multiple operating systems we had in laptops in the past. Ultimately a more optimised experience is where the body is build for the mind and the mind is build for the body.
Nate, thank you for the review and notes. Great content as always! Agent memory is so obvious think - why AI vendors of agent framework creators not implemented it right from the start?... Btw, I like the new style and look of your video - better and more natural. But the slides (possibly generated) are a bit off, and sometimes distract instead of supporting you narrative.
please add that Gemma 4 is almost unusable in openclaw because of the toolcall problems
I tried starting with OC in Feb, March but gave up on it because they were changing it near daily and BRICKING it from even starting
In my vision I saw this architecture too and happy to see you are leading down this route 🎉 one question comes to mind are run times like N8N a possible integration architecture to create activity and function where only a function is needed and not a LLM riducing tocken consumption?
Genius insight. Thanks.
Big Fan. I see you switched to a side view of your desk. Your mic is now DJI instead of the mounted mic. I'm guessing you've upped your YouTube game. Did you buy a teleprompter?
I would love to see an up-to-date comparison with Hermes. Many of us switched from Open Claw to Hermes for its improved memory and reliability... Should I be switching back?
It is beginning to feel like OpenAI is paying Nate now.
Are we just not dealing with the fact the cost of compute is really way more expensive than they are admitting? And the Corporations will not be able to afford the replacement of most of the best human workers, because the AI is so expensive now in its current environment.
If you look at youtube there are many instructions on how to run another model locally with claude, for example gemma4. So why use open claw for the same thing? And instead of one system have two?😊
Thank you Nate for your content, I come from network infrastructure, and was studying cyber sec, until AI happened, Your channel is taking my understanding to an entirely different level. Design the workflow architecture not the replaceable widget. You truly want to just share the knowledge, and it is a beautiful thing.
Maybe your agents are limited to one model, mine aren't and I sure dont use OpenClaw lol
I don’t like your new ai avatar. It feels like I’m watching a claymation
Way too many tech terms used man! Realize that your audience includes non tech people who have vibe coded 70k lines of code in a single app and are struggling with VS and Replit and back to a GPT subscription. Glad I listened long enough to figure what the hell you were trying to say. Still not certain it is safe to use Claw and whether or not it will be able to save me from copy pasting nightmare. BTW…still love you but jeeez
Great video Nate. I moved away from openclaw so I could build a "model-broker" and an "operational memory bus" that can shared across agent harnesses/swarms. Time to try out openclaw again, methinks.
Exactly. Durable workflows and persistent memory matter more than model loyalty. If the runtime holds the logic, the brain can evolve without breaking the system. That is when agent tools move from demos to real production work.
The model is just one piece in your whole setup. You can swap it as you like. If you want to selfhost the minimal stack would be: OC-VM with docker, gitlab, openbao, ansible, openbrain, monitoring, backup. Rules: gitlab first -> IaC/CI, TDD, UAT, only tested results are valid, test results as proof, spec->code->verify against spec->UAT->Sec-Check->Stage->E2E->prod If a step does not work, start from spec with the failure information. Specific Roles with responsibility for tasks or steps in their own sessions, for context clarity. All changes need documentation (Tasks, Issues in gitlab) No work without ticket. Tickets as handoffs. I can and did switch models as I like underneath this setup.
bro your audio