<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo" -->

---
title: Building a New Approach to Enterprise Python Libraries...
description: A Capital One PyCon talk argues that enterprise Python teams should stop building internal SDK wrappers around open-source libraries because AI coding models...
canonical: https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Building a New Approach to Enterprise Python Libraries (Sponsor: Capital One) | daily.dev
og:description: A Capital One PyCon talk argues that enterprise Python teams should stop building internal SDK wrappers around open-source libraries because AI coding models...
og:url: https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo
og:image: https://api.daily.dev/og/posts/d4BBzT1yO.png
og:image:alt: Building a New Approach to Enterprise Python Libraries (Sponsor: Capital One)
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Building a New Approach to Enterprise Python Libraries (Sponsor: Capital One)

**[PyCon US](https://daily.dev/sources/pyconus)** · 63 min read · 0 upvotes · 0 comments

## Summary

A Capital One PyCon talk argues that enterprise Python teams should stop building internal SDK wrappers around open-source libraries because AI coding models are trained on public open-source code, not internal abstractions, and internal wrappers cause hallucinations and maintenance burdens. The speakers propose using transparent middleware (e.g., FastAPI middleware) instead of SDKs, contributing internal patterns back to open source, building dependency graphs and documentation to give AI models context on legacy code, and using agents.md/claude.md files to encode mandatory patterns like correlation IDs and auth middleware. They cite an internal stat that ~40% of Capital One's code is Python and reference an EsoLangBench paper showing LLMs perform poorly on obscure languages, reinforcing that AI works best with well-represented open-source patterns.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://www.youtube.com/watch?v=iRg8fMu_j2w>

## Questions this post answers

### Why do AI coding assistants hallucinate SDK methods that don't exist?

AI models hallucinate SDKs because they were trained primarily on open-source code and documentation, not on internal company libraries. When asked to interact with a vendor API, a model often assumes a matching SDK exists because that pattern appears constantly in its training data, even when no such SDK exists for that specific vendor, leading developers to manually implement authentication and payload serialization instead.

_Anyone debugging AI-hallucinated dependencies can follow ongoing AI coding practices on daily.dev._

### Should enterprises build internal Python SDK wrappers around open-source libraries when using AI coding tools?

Building internal SDK wrappers is discouraged because AI models are not trained on proprietary internal code, so they lack context on how those wrappers work and tend to hallucinate or misuse them. A recommended alternative is using transparent middleware, such as FastAPI middleware, which preserves the original open-source library's structure while adding enterprise-specific behavior like correlation IDs or authentication checks.

_Teams deciding between SDK wrappers and middleware patterns can track these architecture debates on daily.dev._

### How well do large language models handle obscure or low-resource programming languages compared to Python?

Large language models perform poorly on obscure Turing-complete languages, according to the EsoLangBench benchmark by Amit Sharma and Paris Chopra released in March 2026, which tested five rare languages with fewer than a thousand GitHub samples each. All tested models scored in single-digit accuracy percentages, with GPT-52 achieving the best result at only 6% accuracy on one language, aggregated across 80 prompt strategies.

_Developers weighing language choice for AI-assisted projects can follow benchmark findings like this on daily.dev._

## Similar posts on daily.dev

- [Ditching GitHub](https://daily.dev/posts/ditching-github-vdjgxgl6t) · Lobsters · 145 upvotes · 20 comments
- [The AI coding hangover](https://daily.dev/posts/the-ai-coding-hangover-5hylutrto) · InfoWorld · 1 upvotes · 0 comments
- [Enterprise AI Has a Trust Problem. We’re Hearing It Firsthand.](https://daily.dev/posts/enterprise-ai-has-a-trust-problem-we-re-hearing-it-firsthand--hkqefcpop) · Kilo Blog · 3 upvotes · 0 comments

---

Tags: [#python](https://daily.dev/tags/python), [#architecture](https://daily.dev/tags/architecture), [#ai-coding](https://daily.dev/tags/ai-coding)

[View this post on daily.dev](https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Building a New Approach to Enterprise Python Libraries (Sponsor: Capital One)","url":"https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo"},"datePublished":"2026-09-18T22:44:07.492Z","dateModified":"2026-09-18T22:44:36.516Z","description":"A Capital One PyCon talk argues that enterprise Python teams should stop building internal SDK wrappers around open-source libraries because AI coding models...","image":"https://i.ytimg.com/vi/iRg8fMu_j2w/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/iRg8fMu_j2w/sddefault.jpg","isAccessibleForFree":true,"articleSection":"PyCon US","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"PyCon US","logo":"https://media.daily.dev/image/upload/s--DLedNxG1--/f_auto,q_auto/v1777800767/logos/pyconus?_a=BAMAMiWQ0","url":"https://daily.dev/sources/pyconus"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"python,architecture,ai-coding","timeRequired":"PT63M","video":{"@type":"VideoObject","name":"Building a New Approach to Enterprise Python Libraries (Sponsor: Capital One)","description":"A Capital One PyCon talk argues that enterprise Python teams should stop building internal SDK wrappers around open-source libraries because AI coding models...","thumbnailUrl":"https://i.ytimg.com/vi/iRg8fMu_j2w/sddefault.jpg","uploadDate":"2026-09-18T22:44:07.492Z","duration":"PT63M","url":"https://api.daily.dev/r/d4BBzT1yO","embedUrl":"https://www.youtube.com/embed/iRg8fMu_j2w"}}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"PyCon US","item":"https://daily.dev/sources/pyconus"},{"@type":"ListItem","position":3,"name":"Building a New Approach to Enterprise Python Libraries (Sponsor: Capital One)"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/building-a-new-approach-to-enterprise-python-libraries-sponsor-capital-one--d4bbzt1yo#faq","mainEntity":[{"@type":"Question","name":"Why do AI coding assistants hallucinate SDK methods that don't exist?","acceptedAnswer":{"@type":"Answer","text":"AI models hallucinate SDKs because they were trained primarily on open-source code and documentation, not on internal company libraries. When asked to interact with a vendor API, a model often assumes a matching SDK exists because that pattern appears constantly in its training data, even when no such SDK exists for that specific vendor, leading developers to manually implement authentication and payload serialization instead. Anyone debugging AI-hallucinated dependencies can follow ongoing AI coding practices on daily.dev."}},{"@type":"Question","name":"Should enterprises build internal Python SDK wrappers around open-source libraries when using AI coding tools?","acceptedAnswer":{"@type":"Answer","text":"Building internal SDK wrappers is discouraged because AI models are not trained on proprietary internal code, so they lack context on how those wrappers work and tend to hallucinate or misuse them. A recommended alternative is using transparent middleware, such as FastAPI middleware, which preserves the original open-source library's structure while adding enterprise-specific behavior like correlation IDs or authentication checks. Teams deciding between SDK wrappers and middleware patterns can track these architecture debates on daily.dev."}},{"@type":"Question","name":"How well do large language models handle obscure or low-resource programming languages compared to Python?","acceptedAnswer":{"@type":"Answer","text":"Large language models perform poorly on obscure Turing-complete languages, according to the EsoLangBench benchmark by Amit Sharma and Paris Chopra released in March 2026, which tested five rare languages with fewer than a thousand GitHub samples each. All tested models scored in single-digit accuracy percentages, with GPT-52 achieving the best result at only 6% accuracy on one language, aggregated across 80 prompt strategies. Developers weighing language choice for AI-assisted projects can follow benchmark findings like this on daily.dev."}}]}
```

