Skip to content

Threads

A thread is the durable record for one line of research. It keeps the question, the context and decisions that shaped the work, the experiments Silico proposed, and the interpretation that followed in one place.

A thread is more than a chat transcript. It preserves why the investigation started, what changed along the way, and how each result affected the next step. A folder helps organize related work; the thread carries the research context itself.

What a thread contains

A thread can contain:

  • the original research question and follow-up messages
  • files, repositories, datasets, models, and other attached context
  • questions from Silico and the answers you provided
  • plan drafts and revisions
  • launched experiments and their current state
  • figures, diagrams, write-ups, and other artifacts
  • interpretation, competing explanations, and follow-up ideas

Context remains associated with the message where it was introduced. This makes it possible to recover not only which resource Silico used, but why it mattered at that point in the research.

Threads, plans, and experiments

These objects describe different parts of the same research process:

Purpose
Thread Preserves the research question, reasoning, decisions, and continuity between pieces of work.
Plan Describes a proposed experiment before it runs, including its method, inputs, controls, and expected outputs.
Experiment Preserves the execution record and resulting evidence after a plan is launched.

A plan is created and reviewed inside a thread. Launching it creates an experiment, while the plan remains in the thread as the record of what was intended. Results and artifacts from the experiment are linked back into the thread so they can inform the next question or plan.

The thread preserves intent and interpretation. The linked experiment preserves execution and evidence.

Thread history and activity

The conversation is the chronological narrative of the research. It includes ordinary messages as well as decisions, plans, experiment updates, and conclusions.

Thread activity is an index over that history. It surfaces plans, experiments, figures, write-ups, and other important objects so you can move through a long-running thread without searching the full transcript. It does not create a separate copy of the work.

Continue, fork, or start a new thread

Use the same thread when the next question depends on its existing reasoning or evidence. This keeps the research through-line intact and gives Silico the context needed to interpret a follow-up.

Use Fork when you want to branch from a particular Silico response. A fork creates an independent thread in the folder you choose and preserves the conversation, plans, and experiments that existed at that point. Later turns are left behind, and the source thread does not change.

Forking is useful when you want to test a different assumption, pursue an alternative recommendation, or return to an earlier decision without disrupting the original line of work. The Fork action appears beneath a completed Silico response in a thread you own.

Start a new thread when the question can be understood as an independent investigation and does not need the earlier thread's reasoning. You can still attach the earlier thread as context when its evidence is relevant.

Sharing and continuation

Every thread has a live share link. The shared view follows updates to the source while remaining read-only.

Extend this thread starts a new thread with the shared thread attached as context. This is different from an in-thread fork: extending references the shared work as a source, while forking branches your own thread from a specific point in its history.

Sharing does not grant access to credentials, organization settings, repositories, datasets, or compute. See Sharing threads for the complete workflow and access boundary.

See Working with a thread for practical guidance on starting a thread, attaching context, steering the work, and answering decisions.