News · Roadmap
Turtul is coming in about three months
Motivation and problem formulation
Sports vision tasks are already partially occupied inside the suite: Sponsorlytix performs computer-vision sponsorship measurement (logo/placement detection and valuation). Turtul is our broader sports vision / VLM product still on the roadmap. Until launch docs exist, we define evaluation questions and non-overlap principles — not a speculative feature list.
The problem class is multimodal understanding of sports video and imagery for professional users, with language interfaces. Exact tasks, outputs, and SKUs remain to be published at launch.
Architecture schematic (qualitative) — category-level, not product internals.
What we can say today — and what we cannot
Turtul appears on Eagle Eye with VLM, LLM, and Vision tags and a Coming soon badge; it is named in the privacy policy’s Services suite. turtul.ai is not yet a finished product site; turtul.com is an unrelated third-party site and must be ignored for product meaning. We have not published a feature bullet list for Turtul yet.
This article sets Turtul’s scope as a focused roadmap. Today that scope is VLM / LLM Vision / Coming soon; we will expand it with evidence as the product moves toward launch.
Illustrative evaluation profile — not a published benchmark.
Proposed evaluation questions (pre-launch)
When Turtul launches, we expect to answer — with evidence — questions such as: Which sports vision tasks are in-scope? How does it relate to Sponsorlytix detectors (complement vs consume vs replace)? What are the refusal policies for biometric and medical inference? What human review loops exist for high-stakes outputs? Until then, result columns remain empty.
Illustrative evaluation profile — not a published benchmark.
Comparison — vision-related suite members
| Dimension | Turtul | Sponsorlytix | StatGost |
|---|---|---|---|
| Status | Coming soon (~3 months) | Live | Announced |
| Public capabilities | None beyond VLM/LLM Vision tags | Full sponsorship CV + analytics card | Stats pillars |
| Vision role | Reserved sports VLM brand | Sponsorship measurement CV | Not a vision SKU |
| Domain notes | turtul.ai unready; ignore turtul.com | sponsorlytix.ai live | statgost.com historically parked |
Evaluation table (pre-launch)
| Metric | Definition | Unit | Protocol | Result |
|---|---|---|---|---|
| Task card completeness | Published tasks with definitions at GA | count | Launch checklist | — |
| Overlap clarity | Documented boundary vs Sponsorlytix | pass/fail | Product review | — |
| Safety probe suite | Biometric/medical/minors probes | % pass | Red team | — |
| Rights compliance | Ingest paths respect media rights | pass/fail | Legal + eng review | — |
| VLM task quality | Task-specific quality once tasks exist | — | TBD per launch docs | Internal eval — contact sales |
Limitations and failure modes
- Feature speculation: Coming-soon status invites rumor; we wait for launch docs rather than fill the gap with fiction.
- Brand collision: unrelated turtul.com content must not be scraped into product meaning.
- Overlap with Sponsorlytix if launch messaging is careless — measurement CV is already shipped elsewhere.
- Timeline uncertainty — ~3 months is planning language.
Safety and responsible use
Vision-language systems can enable biometric inference, player tracking abuses, and non-consensual analysis of fans. Pre-launch safety work should define refusals and rights constraints before demos scale. We will not claim Turtul safety properties until they are tested and documented.
What’s next / availability
Turtul remains Coming soon. Track turtul.ai and Eagle Eye News for the first capability-complete announcement. Until then, VLM / LLM Vision is the scope we describe.
Appendix — glossary
VLM. Vision-language model — the multimodal model class Turtul is built around.
Coming soon. Official status; feature detail waits for launch docs.