Model library
The models, named
Trained ML models deployed into your application alongside the language agents. Each one below states what it predicts, what it needs as input, and what it cannot do. newc0 publishes no model count — only the models it can show you.
Last updated
| Model | Domain | What it predicts | Status |
|---|---|---|---|
| sales-forecasting | Revenue | Sales forecasting predicts how much of the current pipeline closes in a given period, and how far the total is likely to move either side of that figure. | Deployable |
| lead-scoring | Revenue | Lead scoring predicts which of your open leads are most likely to convert, and ranks them so a limited amount of attention goes to the ones worth it. | Deployable |
Domain · Revenue
sales-forecasting
Sales forecasting predicts how much of the current pipeline closes in a given period, and how far the total is likely to move either side of that figure.
What it needs as input
- Open opportunities with amounts and stages
- How long each has been where it is
- Historical outcomes of comparable opportunities in your own records
What it will not tell you
It reads your pipeline, not your market. A forecast is a distribution with a confidence figure attached, and a founder who treats it as a commitment will be wrong roughly as often as the confidence figure says.
lead-scoring
Lead scoring predicts which of your open leads are most likely to convert, and ranks them so a limited amount of attention goes to the ones worth it.
What it needs as input
- Lead attributes captured at creation — size, sector, source
- Engagement signals recorded against the lead
- Outcomes of previous leads in your own records
What it will not tell you
It ranks; it does not qualify. A high score is a reason to call someone first, not evidence that they will buy, and a cold start with no historical outcomes to learn from ranks poorly until there are some.
Why is there no model count on this page?
Because a count is only worth something if you can read what is behind it. newc0 has confirmed two model domains, so this page names two models. If that list grows to twelve, this page will say twelve and show twelve.
Every platform in this category can print a large number. Almost none of them can put a browsable list under it. That asymmetry is the reason this page exists, and it stops being worth anything the moment the number stops matching the list.
What can a trained model tell you that a language model cannot?
A number that is calibrated against outcomes. A language model can write a persuasive paragraph about which leads look promising. It cannot rank 41 of them by likelihood of converting and attach a confidence figure that means something, because it was never trained to produce one.
This is why both exist in a newc0 company. Language agents do the work that is written. Trained models do the work that is counted. Neither is a substitute for the other, and a platform offering only the first is missing the half that tells you where to spend your morning.
Questions about the models
Do AI agents use machine learning models, or just language models?
- Both, and the difference matters. A language model writes text: outbound copy, a support reply, a draft. A trained ML model produces a number: which lead converts, how much of the pipeline closes. Asking a language model for the second kind of answer gets you a confident sentence, not a prediction.
How many models does newc0 have?
- newc0 does not publish a count. Two model domains are confirmed and both are on this page, named, with what they predict and what they need. A number with nothing browsable behind it is unverifiable, and this site would rather be short than unverifiable.
Are the models specific to my business?
- They are deployed as part of the application newc0 builds you and they read your records — your opportunities, your leads, your outcomes. What a model can tell you is bounded by what your own data contains, which is why a brand-new company gets weaker predictions than one with history.
Can a model decide something on its own?
- No. Model output arrives as a prediction with a confidence figure attached, in the morning report, labeled as model output. It informs what an agent does next and what you look at first. It does not approve a discount, send a refund, or set a price.