How we use AI
Last updated: 2026-10-26
Nettacular is operated by EggBlox LLC. This page says where artificial intelligence (AI) is involved in the AI Model Timeline, what it does with information, and where it is not used.
Summary
AI helped us gather and draft the evidence behind the AI Model Timeline. AI research agents read public web pages, such as provider announcements, pricing pages, model cards and published papers, and proposed records. A person then reviewed every record before it was added to the dataset. You do not talk to an AI on this site, and nothing you do on the site is sent to an AI service.
AI is not used for: answering your questions, generating the forecasts, choosing or ranking what you see, or making any decision about you.
Where AI is used
| Feature | What the AI does | What it receives | Provider (product) | Labelled on the site? | Human review? |
|---|---|---|---|---|---|
| Evidence research and drafting (offline, before publication) | Reads public web pages and proposes records: dates, prices, benchmark results, each with its source link, location in the source and a short quotation | Public web pages and our written research rules. No reader data. | OpenAI (GPT-6 family, through Codex) and Anthropic (Claude, through Claude Code) | Yes: this page, and a note on the AI Model Timeline page | Yes: every record is checked by the dataset tooling, values are re-fetched from their sources to confirm them, and a person accepts or rejects each batch |
The same research can be run with either provider’s agents. Both are companies whose models the timeline compares. See How we keep this neutral.
The forecasts are not AI
Forecasts on the timeline come from ordinary statistical formulas (simple baselines and trend fits over the dated history) that run in your browser and when the site is built. They are not produced by a language model or any other AI model. They are our estimates, described on the timeline page, and they are not statements or plans of any model provider.
How we keep this neutral
- Every record enters the dataset only after human review. AI drafts are never published directly.
- Claims that flatter a provider are held to the same evidence bar as any other claim.
- A value drafted by one provider’s agent that favours that same provider gets an explicit second check, by a person or by the other provider’s agent re-reading the source.
- Disagreements between sources are shown as disagreements. We do not average them away.
- The record of which agent drafted what is kept in our research files and is not shown on the site.
Your data and AI
- Nothing you do here goes to an AI. The timeline is static: your filters, searches, forecast settings and downloads are handled by your browser. There is no chat, no form and no account.
- Training: because no reader content is sent to AI services, none of it is used to train AI models. The research agents receive only public web pages and our own research rules.
- How long it is kept: the AI drafts and their review records are kept in our research files so that corrections can be traced. They contain no reader data.
- Choosing not to use AI: there is nothing to opt out of on the site itself. You can look at any record’s cited source directly from the evidence drawer.
Limitations
AI can misread a table, mix up two model versions or miss something, and can sound sure when it is wrong. Our review catches most of this, not all of it. Check anything important against the linked source. Nothing on the timeline is financial, investment or other professional advice.
Decisions about you
AI does not make decisions with legal or similarly significant effects about you. Nothing on the site profiles you or personalises what you see.
Questions and problems
If you spot a wrong date, price or result, tell us at contact@nettacular.com and include a link to a source we can verify. We correct the dataset and keep the earlier value on record as superseded.
Changes
| Date | Change |
|---|---|
| 2026-10-26 | Page created with the AI Model Timeline. |