Claude project knowledge fills instantly: cap is total context tokens, four workarounds
AOE_234 · reddit · 2026-08-18
A longtime ChatGPT user moving 8 markdown biochem/ochem textbooks into a Claude project hit the knowledge cap after just 7 files. The reason: Claude loads project knowledge into the context window, so the limit is summed tokens across files — not file count or storage quota like ChatGPT; RAG kicks in past a threshold but can't bypass the upload cap.
Workarounds listed: split into two projects by subject; use the Google Drive connector so files stay in the cloud and are searched per query; use a Cowork/desktop project pointed at a local folder (no cap, but phone access requires the computer on); or just use NotebookLM, which is arguably built for this. They also ask for a remote option like ChatGPT's phone-to-desktop connection.
More from Models
- Meta open-sources Muse Spark Open agent model with day-0 TPU support in SGLang — BanghuaZ · 2026-10-16
- Saying "load-bearing" in your prompt makes Opus 5.5 stop holding back — zacharynado · 2026-10-16
- Americans get their Claude accounts banned while on business trips in China — dotey · 2026-10-16
- I priced every Grok Bot call: $20 to $200 plan in 5 days, long jobs eat 52% — jonathanmalkin · 2026-10-16
- NVIDIA releases Nemotron-3.5-Embed-1B, a multimodal embedding model covering 34 languages — jacek2023 · 2026-10-16
- EpochAI pegs GPT-6 training at up to 1.8e27 FLOP; scaling01 calls param estimates way too high — scaling01 · 2026-10-16