Weekly update – 17 July 2026

[ryan]

Greetings from Oslo, Norway. No ProcessWire updates this week as I'm currently traveling. The same will be true next week as well, but after that we'll be back to our regularly scheduled programming. I'm looking forward to getting back and trying out the new Kimi K3 model (in OpenCode Go) -- anyone had a chance to try Kimi K3 yet? It seems to have many of the YouTube AI channels excited. I'm also looking forward to checking out the new PAGEGRID 2.3 from @jploch, which sounds amazing. Have a great weekend and week ahead! 

[wbmnfktr]

3 hours ago, ryan said:

anyone had a chance to try Kimi K3 yet?

Not yet, but a lot of fun with Grok 4.5 which is amazing.

Have a great time in Oslo!

[HMCB]

Safe travels, @ryan.

[wbmnfktr]

I just post this here so you know:

Quote

If you want an extra $5 credit for it use this link and it'll give both of us $5.

That link above is @ryan 's OpenCode Go referral link from the post below - so if you want to try KIMI3 and give back a few OpenCode credits/dollars to Ryan:

 

[Ivan Gretsky]

On 7/17/2026 at 4:42 PM, ryan said:

Greetings from Oslo, Norway

Wow! Excited to be on the same continent with you! Welcome and have fun!)

On 7/17/2026 at 4:42 PM, ryan said:

I'm also looking forward to checking out the new PAGEGRID 2.3 from @jploch, which sounds amazing. Have a great weekend and week ahead! 

What should one do to get this level of promotion 😆 ?! 

[Ivan Gretsky]

On 7/17/2026 at 4:42 PM, ryan said:

I'm looking forward to getting back and trying out the new Kimi K3 model (in OpenCode Go) -- anyone had a chance to try Kimi K3 yet? It seems to have many of the YouTube AI channels excited.

Just tried Kimi K3 with Kilo via Openrouter on a rather small task. But with a lot of custom code and hooks to find and read. It dug deep, tested well, used agent tools like it should and wrote in align with all the rules I gave it. But spent a lot of time. A good one but might be an overkill for easy stuff.

[wbmnfktr]

On 7/19/2026 at 2:15 PM, Ivan Gretsky said:

But spent a lot of time. A good one but might be an overkill for easy stuff.

Same here. A "simple" feature update which would have taken about 5 minutes in total using Grok 4.5 took way longer. Kimi K3 did a lot of things while updating the feature, maybe way too much, and spend over 1,5 hours and burnt through a lot of usage. Maybe perfect for in-depth planning and super weird bugs, but I can't wait 30 minutes for a new fieldset in a module.

[ryan]

I was able to try Kimi K3 for a bit here while traveling. I tested through OpenCode Go. Kimi K3 made some good suggestions for AgentTools, and I had it get started on implementing them. But I found that there seems to be some issue with using it through OpenCode Go, as I'm constantly hitting OpenCode's Kimi usage limits (not my usage limits), so almost everything involves multiple retries. I think it's from OpenCode (the company) being throttled by Kimi (the company). But when it comes to my actual OpenCode usage limits, I really don't get much usage out of Kimi K3 before the 5-hour or weekly limits are exhausted. That's probably because OpenCode charges 2x usage for Kimi K3, while also using a lot of tokens, so it's not very practical to use with OpenCode's Go plan at present. I can also confirm it does seem to take a lot of time to do most things, though uncertain if that's Kimi K3 the model, OpenCode, or just that Kimi K3 is currently very popular and maybe there aren't enough resources dedicated to it. I don't think I'll use Kimi K3 more until it's more practical to use through OpenCode Go, but for what output it did produce, it was very good.