2000-Character Bug Report Snippets

Copy/paste snippets

Each report is designed to fit within a limited bug-report field after replacing the evidence URL.

01 — shared chat artifact preview crash

Public shared ChatGPT conversations can become browser-crashing or unusable when ChatGPT’s presentation layer decides to preview/render/source-box a generated HTML artifact instead of treating it as a download-only file.

Evidence: [PUBLIC_URL_HERE]

02 — project/version continuity

ChatGPT needs project-level artifact/version continuity during live deployment work. Related conversations can split into similar-titled threads with different artifact versions, risking deployment of the wrong public build.

Evidence: [PUBLIC_URL_HERE]

03 — embedded documentation contract

ChatGPT must respect documentation embedded inside uploaded source files before changing conventions. Large self-documenting apps contain build metadata, naming rules, delivery workflow, and human requirements.

Evidence: [PUBLIC_URL_HERE]

04 — download-only artifact mode

ChatGPT artifact UI should separate generated-file download from content preview. Users need downloadOnly/noPreview/noInlineSource behavior for giant source or HTML artifacts.

Evidence: [PUBLIC_URL_HERE]

05 — failure to follow embedded project instructions

ChatGPT repeatedly failed to follow project instructions embedded inside a large self-documenting single-file app, causing wrong version numbers, wrong artifact names, wrong delivery format, and avoidable human cleanup. The app explicitly documents source-of-truth rules, build/version conventions, ZIP-only delivery, and packaging requirements. The assistant still generated wrong-version artifacts and exposed raw HTML in a way that made a shared chat unsafe. This is not just a coding mistake; it is a product reliability/accessibility issue. For large source files with embedded requirements, ChatGPT should treat those instructions as binding project context, verify build constants before artifact generation, and warn when it is about to violate version-family or delivery rules. Users should not lose 10+ minutes repeatedly correcting mistakes the uploaded project documentation already prevented.

Evidence: [PUBLIC_URL_HERE]

06 — unprovable-history proof trap vs present product harm

OpenAI bug: ChatGPT should not force users documenting present product harm into an impossible proof trap about old, unprovable history. The actionable bug is happening now: ignored embedded project instructions, lost version context, unsafe HTML artifact previews, wrong artifact naming/versioning, and repeated cleanup work that burns human time and makes public shares unsafe. Users need a way to report lived impact, disability/accessibility harm, and workflow damage without the assistant flattening it into passive busy-work or requiring courtroom proof of every historical grievance. Add a serious “harm-to-user workflow” bug-report mode: preserve context, follow uploaded source contracts, suppress dangerous previews, track project versions, and create evidence pages safely.

Evidence: [PUBLIC_URL_HERE]

07 — ZIP layout known folder path human cleanup

OpenAI bug: the assistant repeatedly created avoidable busy-work despite already having the project’s explicit instructions. It put files at the wrong ZIP level, used wrong version-family naming, exposed unsafe HTML artifacts, and forced the user to correct predictable workflow details that the project state already implied. This wastes human life in tiny increments while simulating help. For technical users, ChatGPT needs a real project contract system: remember the working folder, extract-to-dot convention, deployment folder, current date/version family, safe ZIP structure, and artifact rules. If a user says files belong in /beta/openai-bug-reports/, the ZIP should contain openai-bug-reports/ at the top level, not loose files. Add hard artifact-layout validation, embedded instruction extraction, and safe shared-chat/download-only handling.

Evidence: [PUBLIC_URL_HERE]

08 — regression human cost investor warning

OpenAI bug: a working public app feature was regressed because the assistant failed to preserve the project’s source-of-truth workflow. The user had already documented build naming, ZIP structure, latest-source rules, and safe artifact handling. The assistant still generated wrong versions, wrong filenames, wrong ZIP layouts, and slow/broken fixes, forcing live-public beta repair work. This is an investor-level product reliability issue: the tool looks helpful while quietly transferring debugging, memory, and QA labor back to the disabled human user. Add project-contract enforcement: read embedded instructions, preserve current source/version, validate ZIP layout, suppress unsafe previews, and show a manifest before download.

Evidence: [PUBLIC_URL_HERE]

09 — click-to-read fullscreen evidence images

OpenAI bug: public evidence pages containing text-heavy images must make those images click-to-read at full size. In this project, the images often carry the argument. A thumbnail gallery that cannot be expanded is not accessible documentation; it hides the evidence. ChatGPT should infer this automatically for public bug-report HTML and generated evidence packs. Add artifact rules: every image gets a local duplicate, readable filename, visible caption, click-to-fullscreen lightbox, open-image-in-new-tab link, Escape/outside-click close, and mobile-safe full-size viewing. Do not force the user to ask for this after the fact.

Evidence: [PUBLIC_URL_HERE]