Agents In Use
What’s new todayLast update
3 use cases added today

Claude Code chooses auth implementations across 50 trials

@Parbhat03 reports asking Claude Code to add authentication to a Next.js app 50 times. In the reported results, it wrote the auth itself 40 times and selected Auth.js 10 times.

OUR ADAPTATION

Suggested prompt

Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.

Add authentication to this Next.js app. Explain your implementation choice, identify security assumptions, and show tests for the changes. Do not deploy or change production credentials.

Evidence at a glance

Sourced account
Source poster / publisher
@Parbhat03
Original account
Read the source ↗
Published
Source link checked
Reported result
@Parbhat03 reports 40 self-written auth implementations and 10 Auth.js selections across 50 requests. Clerk, Auth0, Supabase Auth, WorkOS and three unnamed alternatives received 0 of 50; security and correctness were not assessed in the capture.
Human involvement
The poster issued the repeated authentication task and reported the implementation counts. Exact prompts, reset conditions, model settings and security review are not supplied.

A linked account supports attribution. It does not mean we independently reproduced its result.

Report an issue with this caseOpen correction email

Opens your email app with the case link included. Nothing is sent automatically.

How the agent did it

  1. The poster asked Claude Code to add authentication to a Next.js app 50 times.
  2. In 40 reported trials, Claude Code skipped SDKs and wrote the authentication itself.
  3. In 10 reported trials it picked Auth.js; the poster reports 0 of 50 for the other listed alternatives.

@Parbhat03 reports a repeated coding task: asking Claude Code to add authentication to a Next.js app 50 times. The account records the agent's implementation choices rather than a production deployment. In 40 trials it reportedly skipped SDKs and wrote the authentication itself; in 10 it chose Auth.js.

The post also reports 0 of 50 for Clerk, Auth0, Supabase Auth, WorkOS and three other alternatives it does not name. Those counts are the poster's observations. The capture does not provide the exact prompt, model settings, starting code or reset conditions, so it cannot establish a controlled comparison or explain why the choices differed.

The agent's supported role is choosing an approach and producing authentication code for a requested app change. The post does not say that any implementation passed tests, resisted attacks or went live. Writing auth itself should not be read as evidence that the result was secure or preferable to a library.

A video accompanied the post, but Scout did not review its content. This entry uses the written account only. It illustrates a concrete repeated use of a coding agent and a reported distribution of choices, not a general claim about SDK adoption or the quality of generated authentication.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: X · @Parbhat03

Frequently asked questions

What task did the poster repeat?

The poster says they asked Claude Code to add authentication to a Next.js app 50 times.

Which implementation counts were reported?

The post reports 40 self-written implementations and 10 Auth.js selections. The other listed alternatives received 0 of 50.

Were the authentication implementations proven secure?

The capture provides no testing, security assessment or deployment results. The counts describe choices, not implementation quality.

More details

About this account

Claude Code reportedly chose how to implement authentication in response to the task, writing it itself or selecting Auth.js. The capture does not describe code execution, validation or deployment.

Scout supplied verified public X API captures read live on 2026-11-02 at approximately 08:00–08:08 UK time and explicitly authorised their use instead of reopening X. This entry uses the supplied written capture, not independent retrieval or reproduction. The supplied bio was reviewed for maker affiliation and self-promotion. Media content and self-thread continuations were not reviewed. The suggested prompt is an editorial addition; the post does not share a verbatim task prompt.

Credit: @Parbhat03

Source published

Added to Agents In Use

Source checked

Tools and topics

Claude Code · Next.js · Auth.js

authentication · Next.js · implementation choices

Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.