Sign in

Python Agentic AI

@pythonagenticai.com
9 followers 3 following 16 posts

Agentic engineering for Python developers. Ship AI agents that hold up in production. Free monthly live events + open Slack. Short terminal demos most days. → pythonagenticai.com/bsky

PostsRepliesMedia
Python Agentic AI @pythonagenticai.com · 09/10/2026
A Protocol seam makes the LLM provider a one-line swap. Type your app against a Protocol, not a vendor SDK.  OpenAI, Anthropic and a fake for tests all fit the same slot, and your CLI, bot and web layers never notice the swap. Things you learn in our cohort program -> pythonagenticai.com
000
Python Agentic AI @pythonagenticai.com · 07/10/2026
Pydantic models are at the core of building agents -> pythonagenticai.com
pythonagenticai.com
Python Agentic AI — A community for Python developers building production AI agents
Free monthly live events, an open Slack, and small code-review cohorts for Python developers building production AI agents.
000
Python Agentic AI @pythonagenticai.com · 07/10/2026
Decimal fields keep LLM-extracted money exact. Type the amount as Decimal on the Pydantic model and the LLM's JSON gets validated into exact money at the boundary. Sums come out right, and a third decimal place is rejected before it reaches your database.
120
Python Agentic AI @pythonagenticai.com · 04/10/2026
Agents write code fast. Knowing whether it's right is the slow part. In our cohort you write the code yourself first, learning the patterns behind reliable agents: typed data, Protocols, a service layer, and "the model proposes, you decide". → pythonagenticai.com
020
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 02/10/2026
In an age where we're constantly urged to move faster, sometimes we have to slow down 🏃 💡 So I'm starting a book club: - Read and discuss a classic programming book - 3 weekly threads to discuss learnings + experience - If enough people, get on a call to dive deeper
131
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 03/10/2026
How to deal with probabilistic systems (LLMs)? Build deterministic systems around them. My expense invoice scanner is a good example. The LLM reads the invoice and shows me what it finds. I can accept, edit, or discard the result before anything gets added to my records.
111
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 02/10/2026
Agents write code fast. Knowing whether it's right is the slow part. In our cohort you write the code yourself first, learning the patterns behind reliable agents: typed data, Protocols, a service layer, and "the model proposes, you decide". → pythonagenticai.com
011
Python Agentic AI @pythonagenticai.com · 01/10/2026
Your agent's pending action survives a restart. More → pythonagenticai.com/bsky
000
Python Agentic AI @pythonagenticai.com · 30/09/2026
Retry the agent's tool call. Still one order. More → pythonagenticai.com/bsky
000
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 30/09/2026
This is one of the layers we build in our Python agentic AI cohort: agents that check the model's output against your real data before they act. Starts end of October → pythonagenticai.com
pythonagenticai.com
Python Agentic AI — A community for Python developers building production AI agents
Free monthly live events, an open Slack, and small code-review cohorts for Python developers building production AI agents.
022
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 30/09/2026
User: "Refund order 4817." Model: {"order_id": 4871} Valid schema, wrong order. The refund goes out anyway. 😅 Structured outputs guarantee the shape. Not the truth.
111
Python Agentic AI @pythonagenticai.com · 30/09/2026
Your agent can't pay out a refund by mistake. More → pythonagenticai.com/bsky
000
Python Agentic AI @pythonagenticai.com · 29/09/2026
Never trust the model's JSON. More → pythonagenticai.com/bsky
000
Python Agentic AI @pythonagenticai.com · 29/09/2026
If we read less code how do you validate correctness?
000
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 22/09/2026
When demoing an app, you control the input. A live system has a user doing something random while you're asleep. Better prompting doesn't make your system more robust. It's the reliability layer that's rarely discussed amid the AI hype: - Retries that don't double-fire the webhook
211
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 24/09/2026
In <Python Agentic AI> we have you build an expense classifier agent. Classifying is the easy part: [in] "Coffee at Starbucks $5.50" [out] Food with 95% certainty. You can ship that but the 56th classification it will go wrong. LLMs are unpredictable.
111
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 28/09/2026
With execution speed of LLMs, deep programming fundamentals + knowledge matter now even more. 🚀 And this urges us to keep reading / studying the classics.
121
Python Agentic AI @pythonagenticai.com · 28/09/2026
Juanjo and I distilled these system engineering patterns into a free 14-page guide where we build an expense agent from scratch. 👇 Comment "GUIDE" and I’ll DM you the link! (or grab it at pythonagenticai.com/guide/) #AIEngineering #Python #LLM #SoftwareEngineering #AgenticAI
pythonagenticai.com
Beyond the LLM Call: free guide — Python Agentic AI
Free monthly live events, an open Slack, and small code-review cohorts for Python developers building production AI agents.
000
Python Agentic AI @pythonagenticai.com · 28/09/2026
• Decoupling vendor SDKs with Python protocols • Testing agent logic in milliseconds for free • Type-safe human-in-the-loop fallback patterns • Applying Toyota's jidoka principle to AI automation
pythonagenticai.com
Beyond the LLM Call: free guide — Python Agentic AI
Free monthly live events, an open Slack, and small code-review cohorts for Python developers building production AI agents.
100
Python Agentic AI @pythonagenticai.com · 28/09/2026
Calling an LLM is easy. Building a reliable system around it isn't. Swipe 👈 to see how we tackle real agentic engineering challenges:
100
Python Agentic AI @pythonagenticai.com · 28/09/2026
Test your agent's tools without calling the model. More → pythonagenticai.com/bsky
000
Python Agentic AI @pythonagenticai.com · 27/09/2026
Test your agent's tools without calling the model. More → pythonagenticai.com/bsky
000
Python Agentic AI @pythonagenticai.com · 26/09/2026
Never trust the model's JSON. More → pythonagenticai.com/bsky
000
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 25/09/2026
We're digging into this and more in the next cohort: pythonagenticai.com
pythonagenticai.com
Python Agentic AI — A community for Python developers building production AI agents
Free monthly live events, an open Slack, and small code-review cohorts for Python developers building production AI agents.
021
Reposted by Python Agentic AI
Bob Belderbos @bbelderbos.bsky.social · 25/09/2026
Fail fast: break loudly and early, not quietly and later. A missing API key in production shouldn't become a mystery at 3am. In a recent PR, we moved config into a typed Settings class and made the API key a required field. No key → ValidationError that tells you exactly what's missing.
121
Python Agentic AI @pythonagenticai.com · 25/09/2026
Your agent loop is 11 lines of Python. More → pythonagenticai.com/bsky
010