Mikayla Thompson

[Mikayla Thompson @MikaylaRoseT]

Am I doing loops right?

[Mikayla Thompson @MikaylaRoseT]

@kunchenguid Two stateless navigation and content verification tasks: https://t.co/s9bkiTwjlY

Because we do QA testing, it's heavy on evidence collection, so screenshots for every step (which adds a lot to the time!).

[Mikayla Thompson @MikaylaRoseT]

@kunchenguid A bit lower than the MCP on token usage as well, so it's a promising direction.

[Mikayla Thompson @MikaylaRoseT]

@kunchenguid Axi ended up right on par with playwright MCP. Probably not distinguishable with n=20 runs, would have to collect more data to say anything definitive.

Way ahead of playwright-cli and agent-browser.

[Mikayla Thompson @MikaylaRoseT]

@kunchenguid I'm currently trying Axi in the same tests as my post is based on, and so far it's somewhere in between the CLI and MCP. The requirements ask for a lot of screenshots, and those add a couple seconds every time with Axi.

[Mikayla Thompson @MikaylaRoseT]

@kunchenguid Yes, I'd be very curious what your results are with the playwright mcp! For sure much higher token usage, but maybe competitive on latency.

[Mikayla Thompson @MikaylaRoseT]

@kunchenguid (and playwright-mcp outperformed agent-browser in my tests)

https://t.co/rVnvy3Swvi

[Mikayla Thompson @MikaylaRoseT]

@kunchenguid I ran some benchmarks recently along a similar line--
playwright-mcp vs playwright-cli vs agent-browser
comparing latency/token usage/cost. Any reason you didn't include the playwright options? They seem like the defaults right now.

[Mikayla Thompson @MikaylaRoseT]

Playwright MCP->CLI was not the straightforward swap I expected it to be. I ended up going deep into answering why the token-efficient option was not the performant (or the cheap!) one.

[Josh Ip @joship__]

https://t.co/WeXytmkkeW

[Ranger @RangerNetHQ]

Claude doesn't get to have all the fun using Ranger to verify its work: Ranger Feature Review is available as an @opencode plugin!

More coming soon about our custom integration into the OpenCode Web UI. 👀

[Mikayla Thompson @MikaylaRoseT]

This started as a hackathon project a month ago. We ended up using it every day for our own work and when we wanted to use it on our side projects too, we thought we might as well make it public.

[Ranger @RangerNetHQ]

We don't manually test features anymore.

You can now run continuous QA features in Claude using “Feature Review” by Ranger.

Feature Review is an always-on AI QA that runs in the background, fixing features while providing you and your team with full visibility.

Never ask Claude to fix errors again.

[Ranger @RangerNetHQ]

We don't manually test features anymore.

You can now run continuous QA features in Claude using “Feature Review” by Ranger.

Feature Review is an always-on AI QA that runs in the background, fixing features while providing you and your team with full visibility.

Never ask Claude to fix errors again.

[Mikayla Thompson @MikaylaRoseT]

@scarletinked I transferred off of a team that wasn't a good fit 2 months after joining and my mindset was absolutely "if I can't transfer internally, I'll be leaving the company shortly." Just had 4 years at amazon last week, so I think safe to say that I'm a success story of this change.

[Mikayla Thompson @MikaylaRoseT]

@denverparksrec It appears to still be closed. Is there an updated re-opening date?

[Mikayla Thompson @MikaylaRoseT]

@samsam_washere I LOVE this!

[Mikayla Thompson @MikaylaRoseT]

@samsam_washere I love this so much!

[Mikayla Thompson @MikaylaRoseT]

@samsam_washere That's like Colorado weather! We have a weekly blizzard too, and then some nice spring days in between.

[Mikayla Thompson @MikaylaRoseT]

@samsam_washere Mine does this too! It's terrible and I have no solution.

[Mikayla Thompson @MikaylaRoseT]

@samsam_washere This is incredible!

[Mikayla Thompson @MikaylaRoseT]

@jtannady I recently heard someone use "bus factor" but phrase it as "if X took a bus out of town, what would happen?"