Building an AI agent framework after 14y shipping software taught me:
Most AI agent tutorials show 5-line examples.
Reality? Production agents need:
โ Plan-approve-execute workflows (humans stay in control)
โ Automatic testing + code review before deployment
โ Multi-language support (don't rebuild for each stack)
โ Real error handling (not glossed over in tutorials)
This is what OpenAgents does.
I'm building it in public on YouTube + open-source.
https://t.co/7udLLkl2OW
https://t.co/f1viG8VkMN
Things are changing so fast and one thing I have noticed is personality and capability is being weighed more than ever before.
You need the right people and mindset around you to adapt quickly and learn from each other.
Credentials open the door but no longer seal the deal.
Work on communication and how to show case how you are constantly learning to stay head of the curve.
Be willing to adapt and you will thrive.
AI music has come such a far way in the last year.
We are moving into an era where you can create content easier than before.
I am using this to curate my music, with words that appeal to me. The effect is amazing.
weโve signed Zero Data Retention agreements with all providers for Go
all models now follow a zero-retention policy
your data is not used for training
james has achieved distributed opencode
agents can run on your laptop, on a remote server, in a cloud sandbox provider
shut your laptop and things keep running
open it back up and all the data syncs
delete the sandbox nothing is lost
Surprised it took this long.
We saw what happened with OpenClaw.
We live in a time where models will constantly change.
Finding the ecosystem that serves your purpose is more important than being relevant on one model.
opencode 1.3.0 will no longer autoload the claude max plugin
we did our best to convince anthropic to support developer choice but they sent lawyers
it's your right to access services however you wish but it is also their right to block whoever they want
we can't maintain an official plugin so it's been removed from github and marked deprecated on npm
appreciate our partners at openai, github and gitlab who are going the other direction and supporting developer freedom
I am starting to realise how big the gap is between the people who know how to code with AI and those who donโt.
Most people are just using vibe coding like method of coding which is why I can understand the frustration due to this approach.
Itโs easy when times get busy to โskipโ on the things that keep you going. Donโt sacrifice your health, as โyou donโt have time.โ By skipping it actually reduces your overall energy n actually decreases your available time.
Stay healthy and work hard in blocks!
Been working on understanding different teams requirements and I am noticing many people want to use the same standards but different tools.
For the next release we will be moving to a beta version of managing agent systems between tools for teams.
we've improved our traffic routing for kimi k2.5 (paid version) on zen so it should now be absurdly fast
might hit rate limits here as we scale up but this feels like something different
Huge update, we are rolling out support for ClaudeCode, Cursor and Windsurf for OpenAgentsControl.
More information how this all works to come in a bit!
Personally I havenโt used openspec but the fundamental issue I have with a lot of planning tools like this and spec kit is the AI over engineers tasks & projects and is very verbose. Which too me is too much context.
I will definitely look into this for some insights as i still got a lot of room to improve on the task management.
others also want spec kit so will look at making task management configurable so everyone can use which tool they prefer.
opencode's minimal mode will be radically simple
we already have a full tui so don't need to get fancy - instead it'll feel like ai is just driving your shell
this avoids the glitches you see in agents stuck in the awkward middle
we started by refreshing opencode run
Been super impressed with Kimi 2.5 but not just for coding. The way it can handle multiple tool calls at once has really opened a lot of possibilities.
adding another free model - arcee trinity large
this is the first american open-source model we're offering
it's a solid base model so we're expecting future iterations to be quite competitive
We entering an era where anyone can prompt and build anything. The leverage comes from knowing how to effectively build the right thing with the right reasons.
Donโt off load all your thinking to ai but allow it to do the grunt work for you.