Pasted File Editor
Simon Willison reverse-engineered Claude's file-attachment detection behavior, building a standalone prototype that automatically converts large text pastes into file uploads. The tool also supports direct file opening and drag-and-drop, with image preview thumbnails. This reflects a broader UX pattern emerging across LLM interfaces: treating bulk input as structured attachments rather than inline context, which affects how developers and power users architect prompts and workflows around token efficiency and context window management.
Modelwire context
Analyst takeThe more telling detail here is that Willison had to reverse-engineer Claude's behavior at all. Anthropic hasn't published a spec for how its interface detects and handles large pastes, which means third-party tooling built around that behavior is coupling to undocumented heuristics that could shift without notice.
This connects to a thread running through recent coverage: the gap between what frontier models can handle and what their interfaces actually surface to users. Lovable's report on GPT-5.5 (covered June 1) highlighted a 22% reduction in context loss as a headline win, which only matters if the input pipeline is feeding the model clean, well-scoped context in the first place. Willison's tool is essentially a manual patch for that input layer, compensating for interface behavior that isn't standardized across providers. The fact that a respected independent developer is shipping workarounds for attachment handling suggests the UX layer around context management is still fragmented enough to create real friction for power users.
Watch whether Anthropic formalizes a public API or interface spec for large-paste handling within the next two quarters. If they do, tools like this become redundant; if they don't, expect more third-party wrappers to accumulate around the same undocumented behavior.
Coverage we drew on
- Lovable on How GPT-5.5 Unlocks Better Planning for Complex Builds · OpenAI (YouTube)
This analysis is generated by Modelwire’s editorial layer from our archive and the summary above. It is not a substitute for the original reporting. How we write it.
MentionsSimon Willison · Claude · Codex · Pasted File Editor
Modelwire Editorial
This synthesis and analysis was prepared by the Modelwire editorial team. We use advanced language models to read, ground, and connect the day’s most significant AI developments, providing original strategic context that helps practitioners and leaders stay ahead of the frontier.
Modelwire summarizes, we don’t republish. The full content lives on simonwillison.net. If you’re a publisher and want a different summarization policy for your work, see our takedown page.