Steve Brown

Zscaler Strategy Architect | Agent Builder | Agentic Experience | MCP Apps | Generative UI | Opinions are Mine Alone

[Steve Brown @_SteveBrown]

@mattpocockuk On git, I def check into git. If part of my project likely with its own PRs. One of my experiments is a self building LLM Wiki that’s just a serious of flue agents running as GitHub actions. Haven’t spent a ton of time tweaking it but I may get back to that for a future project.

[Steve Brown @_SteveBrown]

@moshthepitt @mattpocockuk https://t.co/JowRhrEXdM

[Steve Brown @_SteveBrown]

@mattpocockuk This is mostly controlled by my skills so feel free to check them out would love to get thoughts on how they function. I am adding an agent that only does. Knowledge as well but it mostly just relies on the skills. https://t.co/JowRhrEXdM

[Steve Brown @_SteveBrown]

@mattpocockuk I use them in a bunch of ways. So many in fact I built skills so I can easily just build the next one. I actually start a new project by initializing knowledge/ and then feed the LLM docs or links that need to be included.

[Steve Brown @_SteveBrown]

@mattpocockuk Content def grows over time so not only at project launch. Mismatches get cross linked and noted by rule. If the LLM can tell if it is superseded to will add tags to both docs. My skills tell the LLM not to overwrite as many times I want to see the conflicts.

[Steve Brown @_SteveBrown]

@rachelnabors Cool talk. Haven't used the platform yet but will give it a whirl as I explore how to include evals earlier in our product process.

[Steve Brown @_SteveBrown]

@liran_tal Fair, but in those cases I generally type "!code ." or "!cursor ." :)

[Steve Brown @_SteveBrown]

@imrane @rauchg Yeah OKF was really close to what I was already doing so it made sense to adopt something that can be portable

[Steve Brown @_SteveBrown]

@BratDotAI I think we are still in the super early days for all of these models. Yes they are highly capable today and maybe the intelligence levels improve at a slower rate but I believe they will become much more efficient and faster. Cost and speed will be the bar moving forward.

[Steve Brown @_SteveBrown]

I had not worked on Antigravity in a while but my wife needed some help summarizing a large number of lies. Long story short we used her Google Account and I may need to add Antigravity back into my workflows it did a great job.

[Steve Brown @_SteveBrown]

@typecraft_dev I’ve been testing both cmux and herdr. I like them both but interested in hearing why you like herdr?

[Steve Brown @_SteveBrown]

@thdxr I tend to agree but also just because a human won’t interact with my agent doesn’t mean it’s not a better experience for your agent to talk to my agent. Especially if my product has lots of specialized logic.

[Steve Brown @_SteveBrown]

@uday_devops Uh…this is like adding up how much my race car cost. I don’t want to know!

[Steve Brown @_SteveBrown]

Any one in algorithm building pure open source pretty much for fun? Not to build a company but just to share the things you are building?

[Steve Brown @_SteveBrown]

@BratDotAI I’ll we watching for it.

[Steve Brown @_SteveBrown]

@BratDotAI Oh cool app. Is there an MCP so you can effectively share them with agent later?

[Steve Brown @_SteveBrown]

@lawrencecchen This is very cool and could def see this being super valuable.

[Steve Brown @_SteveBrown]

@calcsam For sure!