SYS:READY PIPELINE:SONNET→OPUS FORMAT:IEEE_CONF ● LIVE
SECTION 01 / HERO
v1.0.0
AUTOMATED RESEARCH WRITING SYSTEM

From experiment.py
to publication.

Upload a Python experiment. Receive a complete 10-page IEEE conference paper in under ninety minutes — written, formatted, and ready to compile.

~60 min · turnaround
10 pages · double column
25+ references · IEEE format
2 models · Sonnet + Opus
N S W E
02

The Process

Five sequential batch operations across two model classes. Total elapsed time typically thirty to ninety minutes.

01 INGEST

Upload experiment.py

Submit your Python experiment file via the web console. The system validates encoding and queues the job for processing.

.py file UTF-8 ≤500KB
02 ANALYZE

Sonnet extracts structure

Claude Sonnet reads the source code and extracts the experimental methodology, hyperparameters, datasets, and expected results into structured form.

claude-sonnet-4 batch · 50% off
03 OUTLINE

Construct IEEE skeleton

A complete section-by-section outline is generated with target word counts, equation placements, figure references, and 25+ academic citations.

IEEEtran 9 sections
04 WRITE

Render full LaTeX

Sonnet writes the complete paper in LaTeX with double-column formatting, mathematical notation, tables, figures, and bibliography — ready to compile.

10 pages ~6000 words
05 PARAPHRASE

Opus rewrites — optional

Trigger an Opus rewrite to receive a second version with restructured prose while preserving every technical claim, equation, and citation intact.

claude-opus-4 prose only
06 DELIVER

Email + Download

An email arrives the moment your paper is compiled. Download as LaTeX source for Overleaf or as a fully-rendered PDF directly from the dashboard.

.tex .pdf
03

Capability matrix

What the system delivers, end to end.

A

Full IEEE Conference Format

Every paper compiles cleanly with \documentclass[conference]{IEEEtran}. Double-column layout, proper section hierarchy, equation numbering, and a complete \bibliography section in IEEE style.

B

Batch API · 50% off

Every Anthropic call routes through the Batch API. You sacrifice instant response — you gain half-price processing.

C

Two-tier model strategy

Sonnet handles initial generation. Opus handles paraphrasing — the deeper reasoning model where it counts most.

D

Email notifications

No polling. No refreshing. The system emails you the moment your paper or rewrite is ready.

E

LaTeX source + compiled PDF

Download the .tex file for editing in Overleaf, or grab the rendered PDF directly. Both versions, both stages — original and paraphrased.

F

Dashboard tracking

Every job appears in your console with real-time status. Filter, download, paraphrase, delete — all in one view.

04

The economics

Rough monthly cost for fifteen new papers and fifteen paraphrased rewrites — based on current Anthropic pricing.

Component
Volume / month
Cost
Sonnet · paper generation 3-call pipeline · batch · cached
15 papers
~$2.50
Opus · paraphrasing 2-call pipeline · batch · cached
15 rewrites
~$4.50
Railway · backend hosting Hobby plan · always-on
~$5.00
Total operating cost
30 jobs
≈ $12 / mo
05

Frequently asked

How long does a paper take to generate?

Anthropic's Batch API guarantees results within twenty-four hours, but typical completion is between thirty and ninety minutes. You'll receive an email the moment it's ready.

Can I use generated papers for actual journal submission?

Most reputable venues require disclosure of AI assistance. Treat the output as a strong first draft — review every claim, verify every citation, and disclose AI use per your venue's policy. Direct submission without review or disclosure is academically dishonest and frequently constitutes a publication ethics violation.

What happens to my Python code?

Stored in a private PostgreSQL database accessible only to you. Sent to Anthropic's API for processing under their data retention policy. Never shared, never resold.

Why both Sonnet and Opus?

Sonnet generates strong first-pass papers at lower cost. Opus excels at deep reasoning — exactly what paraphrasing requires when you need substantial restructuring without losing technical fidelity.

What if PDF compilation fails?

Download the LaTeX source instead and upload to Overleaf — it handles every edge case the embedded compiler might trip on.

[ READY ]

Upload your first experiment.

The console accepts any valid Python file. Drag, drop, wait, receive.

Open Console