Natan Rivera
299 posts

Natan Rivera
@natan_RIAL
Mech Engineer (Rice) 🇸🇻 | Data centers by day, shipping products by night | https://t.co/7GUvuHHWZ9 $150 MRR | https://t.co/Qt1sjbiRqT $0 MRR
Katılım Ekim 2011
105 Takip Edilen86 Takipçiler


Today I got the best feedback from a user. “Would prefer to have my workout ready instead of needing to talk to AI every time” also “Would love that remembers exactly what workout did I do.
Even with context loops for the agent the non deterministic part always changes the results a bit. But letting AI build a workout every time, it’s very token intensive, but I know exactly what to do to avoid this.
English

@hrprtsingh9 The approval is quick just 2 clicks, check the proposal, then approve and it’s ready. But proposing does take a while since the agent have to run a series of tools for context. I’m still trying to find the right balance and handle the non-deterministic part.
English

@natan_RIAL Human-in-the-loop for workout plans is smart. Nobody trusts AI with their body yet. How fast is the approval flow — does it feel like friction or safety?
English

Here is an example of a human-in-the-loop agent I built for my app. The agent would run tools to get user context, and then propose a workout. User needs to approve or requests edits. The workout proposal is part of my state, which follows the workout model type so I can reuse in UI.
If approve, I run a tool to write in DB and create workout for the user. Since the proposal is part of state, the agent can iterate on it as long as the graph loop is open.
English

@sydneyrunkle @hwchase17 Great post. When would you recommend middleware over just building directly in LangGraph? For those of us already comfortable with graphs, what's a good reason to switch?
English

Just made my app freemium to see how it goes. A lot of people liked the features but not everybody was willing to add their credit card right away mostly in Latin America.
Of course the AI coaching feature is still paid but I made common GPS, lift tracking, Apple Watch integration free up to a certain point.
I got 3 trials last night so will see.



English

If you’re a developer or solo founder, I want to connect.
Let’s #connect👇
English

@jakobjelling Workaapp.io an AI fitness coach with GPS, lift, guided circuits, BLE and Apple Watch integration
English

@lukatofocus @ygg0f I haven’t I’ve been looking into it. Right now I’m directly reaching out to people I know. Will see how it goes lol follow to see if it works out. I’ll follow back
English

@natan_RIAL @ygg0f thats the right call. at $150 MRR youre past the validation stage. curious whether distribution is a content play or outreach for you - have you tried reddit communities for hybrid athletes?
English

Ask a bunch of questions, plan, let it rip
Garry Tan@garrytan
You can one-shot anything if you plan well before, and test well afterwards.
English

How many people is using @LangChain langsmith to Eval their agents? It’s pretty good UI. I used the @OpenAIDevs openAI API rn, and I just noticed a couple improvements thanks to this.

English

@Twarga_Dev Yeah mostly with social media brain rotting us every day
English

@natan_RIAL Reading is a lifelong habit of mine that I'll never give up. It’s like a workout for the mind
English

What do you think of this langgraph agent. I made a carrousel generator for personal brands at the link below. Really curious if people want image generation or App Store marketing images.
pixel-carrousel-frontend.vercel.app

English







