Max Shaw

Building @trywindmill. Built @GifJifApp. Formerly Product @Yext,

[Max Shaw @_maxdshaw]

@SepasZohari It all funnels into a single list prioritized by a me.

Usually real feedback is going to be high priority then agent feedback but no hard and fast rules.

Simulated feedback is usually most helpful pre-launch.

And agents are incredible at finding bugs before your users do.

[Max Shaw @_maxdshaw]

The more feedback you can get flowing in to your product the better.

I wouldn't recommend fully automating the build process unless the app is just for you but you need a fast loop from feedback -> improvement. We always have a human in the loop to make sure we are building the right thing.

For feedback we use a combination of a bunch of methods:
* Users can leave feedback via our agent
* Users can leave feedback in the platform
* Our agent can leave feedback based on self reflection
* We have simulated user feedback agents which send in feedback.

[Patrick Collison @patrickc]

I'm now building an "improvements" feature into all of my personal vibecoded applications: a freeform text input that's hooked up to an agent that yolo implements whatever is submitted.

It's an interesting experience -- the activation energy barrier to sanding any rough edge or implementing any new feature becomes extremely low. Someone else who uses one of the applications I created has already implemented 150 improvements over the course of a few days. This pattern feels like a new kind of tool -- liquid software? -- and every application that doesn't have this feature now has a slightly rigid and anachronistic flavor. Why can't I just tell it how I want it to be different?

Now, it's not obvious to me how this pattern can or should scale. Does it become unmaintainable after a while? (Maybe model intelligence improvement outstrips my feature appetite such that it's fine?) How does collaborative software work in this model? Can software designed this way be secure and reliable?

Whatever the larger questions about broader deployment, it definitely works great in a personal context. It reminds me a little of the visions of Lisp/Lisp machines, Smalltalk, and HyperCard -- systems that the user can take control of and modify. Those visions never quite worked, though: the vast majority of software today is not end-user modifiable in any useful way. So, while this "liquid" affordance is a very small touch, the way it changes one's relation to the software feels a bit transgressive and empowering in a Lickliderian and Engelbartian way.

[Max Shaw @_maxdshaw]

@thdxr I much prefer a dedicated browser for agents vs. taking over chrome. Did you consider that?

[Max Shaw @_maxdshaw]

@muzzdotdev @greptile graphite

[Max Shaw @_maxdshaw]

@greptile My first ever 0/5!! What do I win?

[Max Shaw @_maxdshaw]

@omooretweets Impressive growth! What's the y-axis?

[nicole alonso @_nicolealonso]

A few months ago, I tweeted asking why more companies don't sponsor marathon runners.

This weekend, we got to do exactly that at @trywindmill.

We sponsored 6 NYC Marathon runners with custom race-day shoes, and seeing them cross the finish line with the Windmill logo flashing by was an incredible moment.

We spend so much time celebrating what people accomplish at work. Why not celebrate what they accomplish outside of it, too?

Performance is what we think about all day at Windmill. On Sunday, we got to watch high performance in person.

[Max Shaw @_maxdshaw]

@NickADobos @jackfriks Same thing herer. Have a multi-media skill for image/video/audio etc. So much better then trying to sign up for each model.

[Philip Sørensen @philipxyz]

@trywindmill = high performance (reviews)

[Max Shaw @_maxdshaw]

@e2e software verification is so important these days.

I'm curious though why is this better then a simple prompt - I think agentic QA excels precisely because it's super open ended and agents are good at finding stuff.

I worry you lose that with some sort of "language".

[Max Shaw @_maxdshaw]

Today Windmill has

@_nicolealonso running 100km
@emyang___ running 26 miles

And I might run 3 on the treadmill watching football.

What a day for peak performance

[Andrew Yeung @andruyeung]

100 km and 15.5 consecutive hours later… @_nicolealonso finished!

[Max Shaw @_maxdshaw]

@WisprFlow Is this expected? I've barely used it this week. Is this a new limit?

@granola whats your limit?

[Max Shaw @_maxdshaw]

@bentossell @meetgranola @turbopuffer @pierrecomputer @linear @sfcompute I’ll throw our hat in the ring @trywindmill

[Mark Tanner @TarkManner]

nothing better holiday delights for @trywindmill customers, great stuff @bennyautomatic

[Alfred Lin @Alfred_Lin]

As we get into performance cycle season, remember you need to do both backward-looking and forward-looking reviews.

Most high-performing companies manage the review *process* well. But the reviews are typically backward-looking and used only to determine compensation. The review process is delivered, and the person stops listening after you give them their number.

You also need to have a forward-looking review. This includes career development but also expectations for the coming year, including success metrics, goals, and plans that fit into the operating planning cycle. Each person's forward-looking career development plan should be as detailed as their backward-looking review.

Ensure you're giving feedback continuously throughout the year. Talking once a year or quarter isn't enough to separate your high performers from your low performers, and correct course. But also make the most of the natural reflection that happens at the end of the year by looking both backwards and forwards.

[Max Shaw @_maxdshaw]

@DanielLockyer I found my new guy https://t.co/KNp5IP7Yfq

[Max Shaw @_maxdshaw]

@larsencc Why do you care about code comments? We gave up that agent fight

[Mark Batsiyan @mbatny]

@trywindmill is shipping like crazy. Their new agentic onboarding is already loved by many customers and is now out in the wild to become the go to solution for the best scaling companies to onboard new employees.

Just another step for @_maxdshaw, @bdistel, and @TarkManner as they build a generational company at Windmill.

[Max Shaw @_maxdshaw]

@emilkowalski Do you read the code your agents write anymore?

[Harsh @0xharsh_]

today we’re launching agentic onboarding at @trywindmill

it’s the natural next step toward our goal of helping people succeed at work

the average new hire takes 8 months to become fully productive. with windmill, it’s done 3x faster

before day one, Windy, our AI agent, builds a ramp plan from what the manager and People team already know about the role. then it runs that plan in slack

it brings people what they need when they need it and answers questions with company context. when someone gets stuck, it adjusts the plan accordingly

we're currently live with amazing customers like @Etched and it’s incredible to see the results