Research log
From the first release to an independent final test.
The log automatically records key processing milestones. Research conclusions are published only after extraction, validation and replication are complete.
Current progress
Loading status
— releases processed. Last public update: —.
Current phase: —
Analysis started
The fixed extraction schema, Qwen3-4B LLM and resumable storage have been activated.
Original dataset complete
All 607 releases from the initial thirty companies have been reanalysed.
One quarter of the full corpus
The independent replication phase is well under way.
Halfway point
At least 3,247 of the 6,493 releases have been processed.
Three quarters processed
Processing is approaching the final replication batches.
Extraction complete
Manual validation, model comparison and the independent final test follow next.
What happens after 6,493?
Validation sample
A blinded human assessment of releases sampled in advance.
Model comparison
A price-based baseline, the original text features and Qwen features evaluated on the same data splits.
Public conclusion
Results, uncertainty, examples and limitations are published together.