ἵστωρ Download

Istor and ChatGPT, compared.

Both answer questions. One is a general-purpose assistant that runs in the cloud. The other is a research instrument that runs on your desk and checks its own claims.

ChatGPT, by OpenAI, is a general-purpose cloud assistant: frontier-scale models behind an account, and files you attach are uploaded and processed on OpenAI's servers. Istor is a local research notebook: documents never leave your machine, every claim in an answer is verified against the passage it cites, and it refuses when your sources do not contain the answer. ChatGPT is the broader mind. Istor is the one that shows its work.

What ChatGPT does well.

ChatGPT is the most widely used AI assistant there is, and the reasons are real. Its models are frontier-scale, far larger than anything that fits on a consumer GPU, and that shows in general knowledge, difficult reasoning, and fluent writing across nearly any topic. It accepts file uploads in a conversation, can read the web when asked, and carries a huge ecosystem of integrations. As a general-purpose helper for drafting, brainstorming and quick questions, it is the default for good reasons.

The honest comparison starts where those strengths stop being the point: with documents you cannot upload, answers you intend to check, and work that has to run without a server.

Where Istor differs.

Istor is narrower on purpose, and the narrowing shows up in three places.

The first is where the documents go. In ChatGPT, an attached file is uploaded and processed on OpenAI's infrastructure under OpenAI's terms; whatever the policy says, the copy now exists on a server. In Istor the document never moves. Extraction, indexing, retrieval and inference all happen on your machine, so there is no upload to weigh and no policy to trust, and unpublished work stays unpublished.

The second is the gate. ChatGPT writes confident answers and, when sources are involved, can name them, but the claims are not checked against those sources before you read the answer. Istor runs a grounding pass on every claim in a draft answer, verifying it against the specific passage it cites, and refuses rather than guess when the library does not contain the answer. For research, "your sources do not say" is a finding, not a failure. Why chatbots invent citations in the first place is the background to that difference.

The third is the desk. ChatGPT is a chat window in a browser behind a login. Istor is a notebook on your machine: sources on one side, answers with citations beside them, notes living next to the passages they came from. It works on a plane, in a dead zone, and on hardware as modest as a GTX 1650, with no account anywhere in the loop.

Side by side.

IstorChatGPT
Where it runsYour machine, fullyOpenAI's cloud
DocumentsStay on your diskUploaded to OpenAI's servers
AccountNoneOpenAI account required
Model scaleCurated small models, GTX 1650 class GPUFrontier-scale, datacenter hardware
GroundingHard gate: every claim verified, refusal when sources lack the answerCan name sources; claims are not verified per passage
Works offlineYes, over your own libraryNo
NotesBuilt in, beside the sourcesNo
PriceFree, open sourceFree tier, paid plans above it

Which one fits.

Use ChatGPT for everything a general assistant is for: drafting, brainstorming, quick questions, wide knowledge where the material is not sensitive. Nothing local matches a frontier model's raw range, and this page does not claim otherwise.

Use Istor when the work is research over sources: papers, reports, transcripts, scanned archives. When the material is private or unpublished, when every claim in the answer needs a citation you can open, and when the work must continue offline, a grounded local loop is not the budget option. It is the instrument built for the job.

The full picture of Istor is on the home page.