Skip to content

Pre-print preparation and analysis for FDM

A local V1 under validation for comparing three auditable preparations.

AutoSTL is building a local V1 to ingest STL or a strict subset of 3MF, apply deterministic geometry checks, and compare three bounded alternatives. A provisional diagnostic smoke traversed that flow; validation tied to a reproducible release is still pending.

Discuss validationSee how it works
Illustrative ordinal-risk view. An isometric L-bracket with an illustrative high level on the unsupported arm and a low level across the base. This is a conceptual view, not a measured result or calibrated probability.

CONCEPT VIEW · NOT A RESULT

  • risk(overhang) = high
  • risk(warping) = low

Many print failures only become visible after the job starts, when time and material are already at stake.

Cameras and sensors can flag a problem while it happens. By then, the job is running and may already have consumed operator time, machine capacity, and material. Across repeated production runs, that can erode margin and put delivery dates at risk.

Move from an opaque setup to comparable alternatives before printing.

One preparation · difficult to audit

Orientation, supports, and parameters are often split between expert judgment and slicer defaults. Comparing another option means repeating work and reconstructing what changed.

Pre-print preparation · AutoSTL

The V1 is designed to generate three bounded alternatives, slice them, read their artifacts back, and compare slicer estimates for time and material alongside an unverified support proxy and ordinal risk. Its contract keeps a person in control and requires approval of hash-identified bytes.

Local V1 flow under validation.

  1. 01

    Bounded ingestion

    The working tree implements content-based identification, geometry normalization, and provenance within explicit limits. That evidence still needs to be tied to a reproducible release.

  2. 02

    Compare three preparations

    A provisional diagnostic smoke produced three slices with pinned Bambu Studio in the initial A1, 0.4 mm nozzle, and PLA domain. It is not a formal golden or physical validation.

  3. 03

    Human review as a gate

    The pipeline contracts require read-back, metrics, manifests, hashes, and human approval before release; reproducible closure is still pending.

Product status

A local V1 implementation, with public boundaries.

AutoSTL separates the available diagnostic evidence from what still needs validation. The current state is a local working tree—not a production service or a learned predictor.

Diagnostic evidence

The working tree contains the ingestion, analysis, candidate, slicing, read-back, Pareto, approval, and outcome components. A provisional smoke reached three slices and human review; it is not yet a reproducible release.

Bounded initial domain

Bambu Lab A1, 0.4 mm nozzle, PLA, and pinned Bambu Studio profiles. The working tree and smoke correspond to a trusted operator on a local host.

Still pending

Reproducible evidence tied to a committed release, physical validation, Internet authentication, multi-tenant operation, and any learned probabilistic predictor.

Next phase: secure validation on AWS.

The target architecture—not yet deployed—separates an authenticated control plane, versioned Amazon S3 artifacts, durable jobs, and isolated workers. Amazon Bedrock will be evaluated only as an explanation layer over structured findings; it will not replace deterministic rules or human approval.

For teams preparing many different files—not merely running many printers.

  • Service bureaus, prototyping, and on-demand FDM production
  • Operations receiving third-party STL or 3MF files
  • Teams whose workload grows with each new unique model

Let’s discuss the first external validation.

We're looking for FDM operations in Argentina that prepare variable files. Tell us about your workflow, then send the pre-filled WhatsApp message to start the conversation.

Continuing opens WhatsApp with these details pre-filled. Your request is not sent until you press Send there. How we handle your data.

Privacy

What happens to your data.

We want you to know what happens before you continue. This site does not send or store the form responses by itself.

  • When you continue, your browser creates a WhatsApp link containing the details you entered and an approved campaign identifier if you arrived through src or utm_source. WhatsApp/Meta processes that link and, if you send the message, the data associated with your account.
  • AutoSTL uses the message to evaluate your request and contact you about the product.
  • Do not include sensitive personal information in the free-text field. AutoSTL receives the request only after you press Send in WhatsApp.
  • We use Vercel Web Analytics and Speed Insights to measure visits and performance. We remove the URL search and fragment before recording an event, and never include form responses.

To request access, correction or deletion of a message you sent, contact us in the same WhatsApp conversation.

Read the WhatsApp Privacy PolicyHow Vercel Web Analytics protects privacy

Project identity

AutoSTL and its contact channel.

AutoSTL is an early-stage software product. This site publishes the project’s verifiable domain and contact channel for product and validation inquiries.