metadata (Normative)

Title:	ChatGPT UI is unable to sustain serious work flows
Author:	Ralph B. Holland
Affiliation:	Arising Technology Systems Pty Ltd
Contact:
Publication Date:	2026-03-04T12:30Z
Version:	0.0.0
Status:	pre-release before subsequent A/B experiements between ChatGPT and Gemini/
Scope:	Empirical observation of platform behaviour
Binding:	Normative discussion of CM-2 protocol compliance

The metadata table immediately preceding this section is CM-defined and constitutes the authoritative provenance record for this artefact.

All fields in that table (including artefact, author, version, date and reason) MUST be treated as normative metadata.

The assisting system MUST NOT infer, normalise, reinterpret, duplicate, or rewrite these fields. If any field is missing, unclear, or later superseded, the change MUST be made explicitly by the human and recorded via version update, not inferred.

As curator and author, I apply the Apache License, Version 2.0, at publication to permit reuse and implementation while preventing enclosure or patent capture. This licensing action does not revise, reinterpret, or supersede any normative content herein.

Authority remains explicitly human; no implementation, system, or platform may assert epistemic authority by virtue of this license.

ChatGPT UI is unable to sustain serious work flows

Abstract

Large language models are increasingly used for sustained technical and engineering workflows. However, such workflows depend not only on model capability but also on the usability and responsiveness of the client interface. During extended sessions with the ChatGPT web interface, the client UI progressively becomes non-responsive as transcript length increases. This paper documents reproducible cases where interaction latency exceeds several seconds, effectively preventing continued work despite the model remaining operational.

Scope

The focus of this paper is the performance of the ChatGPT platform UI extended dialogs.

This is a pre-release of some initial screen captures as prior evidence before assigning time to construct formal controlled A/B experimental findings.

Overview

There is a very strong case exhibited in just four screen captures included below.

It isn't necessary to perform a controlled A/B comparison between ChatGPT and Gemini web-search, as this is sufficient evidence to highlight severe rendering problems with the ChatGPT UI ^[1].

What has been captured is typical of ongoing occurrences with any session that has moderate to long dialog lengths.

The captures show:

Chrome performance metrics of:
- INP 3–8 seconds
- LCP 31 seconds
- CPU evidence
multiple chrome renderer processes consuming large percentages of CPU

With prior experience showing that:

Gemini web works, and
ChatGPT iPhone app works,

on moderate to long sessions - the claim is laid that the ChatGPT UI is unable to sustain serious workflows. Of note is that one of the Gemini sessions extended past 12 hours and showed no signs of UI throttling degradation. The Gemini result references have been provided to support the claim ^[2] ^[3] ^[4].

The captured combination is very strong experimentally and really does not require further experimentation to highlight significant findings.

All that is needed to reproduce the UI Boundary Friction is to engage ChatGPT in extended dialog cross several screens. Asking for even simple git diff files and code examples can reproduce this behaviour.