Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are live.
Available on Pro, Max and Team plans
Gemini 3.6 Flash:
🔹 Better knowledge work & multimodal performance. 17% fewer output
Gemini 3.5 Flash-Lite:
🔹 Fast, most cost effective, 350 output tokens/second
Try now with /model in the Command Code CLI. 
nice! glad to hear it. i worked on tilemill+mapbox studio in a previous life and frank am surprised at how little mapbox has "pivoted to ai" so far 😂. mapbox studio is the kind of power-user application that project managers are always hungry to dumb down.
Google's Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are now available in Netlify AI Gateway and Agent Runners.
Zero configuration. No API keys to manage. Just call the model from your Functions and go. www.netlify.com/changelog/ge... 
haha oh no, are they doing a vibecoding version of mapbox studio?
they're also exploring new territories of low contrast, the reinforcement learning once this generation of slop gets in the training set inexorably leads to black-on-black design
Nine new Next.js CVEs: SSRF, a middleware auth bypass, DoS, and cache confusion. Fixed in 15.5.21 and 16.2.11.
Good news for Netlify users: three of them don't affect Netlify-hosted sites at all. The per-issue breakdown, plus what to do now: www.netlify.com/changelog/20... 
ACP v2 is available as a draft. 📝
Highlights:
- Updates beyond user-initiated turns
- Stream & update messages, tool calls, and terminal output
- Structured file changes and better diffs
- Improved extensibility and forward compatibility
Help shape v2:
agentclientprotocol.com/announcement... building a slop gtm collection www.are.na/tom-macwrigh... 
Post for Me scaled a social media API to $13K MRR in under a year on Unkey.
Keys from day one, then they moved hosting over too and deleted Render, Railway + Vercel.
Auth + deployments, one platform.
unkey.com/case-studies... totally agree, i've got a big stash of gels and such, this is just for variety and multi-day trips!
getting into endurance sports has put me at a point where i have to type in "squeezable peanut butter NOT for dogs" in the product search bar
Command Code is the most used coding agent on open router overtaking Claude Code, Cursor, OpenClaw, Kilo, Cline, etc in just two months.
Command Code is purpose built for open models. Devs using it to get the most out of open models, save the most money, get the best caching & tool call repairs 
The API and MCP product blog is here elicit.com/blog/the-el... We published the full evaluation including code, data, prompts, and our outputs, so you can run it yourself: github.com/elicit/elic... We rebuilt Elicit's search from the ground up and tested whether it actually was better.
On BioASQ benchmark, Elicit had the highest recall at every result depth we tested from 10 to 200 results.
Read the full write-up: elicit.com/blog/evalua... 
the models seem to have very little respect for WCAG color contrast compliance
@aphera.co enjoyed playing around with the beta! now that it's out, i'm tempted to switch from capture one but my biggest q is whether aphera is aiming to have more photo organization primitives (search by time, tags, face recognition etc) or if there's a companion app people use? 
Expect practical talks, pizza, drinks, and good conversations with Berlin’s TypeScript community.
📍 Prisma Berlin office
🕡 Doors open at 18:30
Join us:
pris.ly/ts-meetup-july This time, the theme is AI-assisted TypeScript development: how developers & coding agents can build, test, review & maintain TypeScript apps together.
We’re also looking for speakers with practical experience & lessons worth sharing.
Submit a talk:
pris.ly/typescript-... 🇩🇪 The last TypeScript Berlin meetup at the Prisma office was packed. We explored architecture in TypeScript, shared practical lessons, and met plenty of developers from the local community.
We’re back on Friday, 31 July. 
Our engineering and design team has been testing 20+ side-by-side comparisons across frontier and open models. All runs are public and open source. Benchmark for this demo here:
github.com/CommandCode... Tested Kimi K3 to build a 3D portfolio website.
- Used the /design command.
- Built in 3 prompts: build, continue, then recolor
- Full session cost: $0.57
The interesting part: Recolor replaced and added an entirely new palette with the same layout and animations in seconds. 
Our engineering and design team has been testing 20+ side-by-side comparisons across frontier and open models. All runs are public and open source. Benchmark for this demo here:
github.com/CommandCode...