Load It
Once.
Aleph keeps big working context in RAM so agents can search, run Python, recurse, and return small answers instead of re-sending the whole file.
Install the Codex CLI, run aleph-rlm install, and let Aleph wire the default shared-session sub-query setup for you.
The Loop
Load once. Search, compute, recurse, finalize.
Work On Data,
Not Prompts.
Aleph stores the full working set in a Python process, then gives the model tools for targeted inspection and computation instead of repeated paste-and-pray prompting.
- + load_file — Put repos, logs, and docs into Aleph memory
- ⌕ search_context — Find the relevant slice first
- <> exec_python — Compute against the whole context
- ◎ save_session — Resume long investigations later
SESSION TRACE
STATE: READY TO FINALIZE
Small answer out, full context still in memory
Recursive By
Design.
Aleph is shaped by Recursive Language Models: keep the working set outside the prompt, let the model search and compute over it, and recurse only when one pass is not enough.
"Each iteration lets the model refine its search based on what it learned, rather than betting everything on one attention pass."
"What my eyes beheld was simultaneous, but what I shall now write down will be successive, because language is successive."