Viviano

🤖AI Product @WHOOP | 🐀 Tracker of Rats | 🏈 Commissioner | 🧠

[Viviano @v_cantu]

@PabloAntonio You should see our hackathons

[Max Weinbach @mweinbach]

Whoop Coach is great

[Alex Cohen @anothercohen]

Just picked up the new Fitbit Air

[WHOOP @WHOOP]

Logging your daily habits should help you understand yourself better… but can often feel tedious, repetitive, and easy to skip.

That’s exactly what we’re fixing in our latest product update. Keep reading to find out what’s new. 👇

[Viviano @v_cantu]

Ok I take it back.
I have no idea what I’m doing

[Viviano @v_cantu]

I think i'm figuring out this SEO thing

[Viviano @v_cantu]

Read more about it here:
https://t.co/k2XYGp7q88

[Viviano @v_cantu]

We created our own language for writing prompts.

We call it HPML (Hyper Prompt Markup Language). It's not just a templating language it's integrated with our tool calling system. So the same tools available to our agents can be called inline like:

{{fetch_weather(city="Boston")}}

Being able to declaratively pull in any data you need into exactly the right place in a prompt matters a lot for @WHOOP since our agent workloads are very reliant on data.

For example, say we have an agent that needs to analyze a workout. We could give the agent a fetch_workout tool and tell the agent to call it. This lets us reuse our data querying infrastructure, but it's not guaranteed and it adds extra latency.

Instead we can just write an HPML prompt and inject the data directly like "{{fetch_workout(id=workout_id)}}"

The symmetry of dynamic and inline tools lets us focus our integration efforts on just tools. If a tool exists we can use it everywhere.

[Viviano @v_cantu]

It ain’t much but it’s honest work

[Viviano @v_cantu]

@rtwlz @levelsio Can you have it play Oscars music if a meeting goes over

[Viviano @v_cantu]

@elliott__potter Congrats! Awesome product and awesome team!

[Viviano @v_cantu]

It continues 🚀

[Viviano @v_cantu]

I think i'm figuring out this SEO thing

[Viviano @v_cantu]

I guess we’re going back to MCPs

[Thariq @trq212]

This is a research preview that we'll be expanding more on.

Read more in our docs on Channels here: https://t.co/apaPL9496Z

[Viviano @v_cantu]

I need this for my daughters toys

[Massimo @Rainmaker1973]

The Robotics team from Wissahickon High School in Ambler, Pennsylvania, built the robot Miss Daisy XXIV that picks up balls and shoots them into a container.

[Viviano @v_cantu]

You have to pick your tradeoff:
- higher cost
- higher latency
- worse accuracy

....or wait until OpenAI ships a new mini model

[OpenAI @OpenAI]

GPT-5.4 mini is available today in ChatGPT, Codex, and the API.

Optimized for coding, computer use, multimodal understanding, and subagents. And it’s 2x faster than GPT-5 mini.

https://t.co/DKh2cC5S3F

[Viviano @v_cantu]

@wesbos https://t.co/ovOiIRPCNa is just a marketing site, it’s not like https://t.co/Dvl2xb2zC7

[Viviano @v_cantu]

No thanks

[Temu @shoptemu]

Order your snacks right here on Temu – what are you waiting forOrder your snacks right here on Temu – what are you waiting for?

[Viviano @v_cantu]

@YifanBTH It’s kind of both. If those tokens were during their highest traffic periods then you are getting your money’s worth. But if those tokens were during low traffic then you’d be overpaying.
In a way the rate limit model is the best for everyone

[Viviano @v_cantu]

It's so easy to spot things, I might do this to preview all my changes

[Viviano @v_cantu]

Asked claude to migrate my entire site from gatsby to next. overnight. Then asked it to build a comparison view to audit its own work. Tech debt is so easy now.