Takeaways

#
  • Trainer argued that reliable domain data and subject-matter expertise were central to useful investment AI.
  • The demonstrated New Constructs agent was in beta and used the firm's fundamental and valuation methodology.
  • The hosts clarified that a ranking of the best stocks did not mean a ranking of technical price patterns.
  • Trainer separated his long-term preference for Google from his assessment of that week's earnings expectations.

David Trainer wanted to change the question investors were asking about AI. On the October 27, 2025 Morning Show, the New Constructs CEO argued that spending more on computing power did not by itself create useful financial research. The harder task was giving a system reliable data and the expertise needed to interpret it.

JC Parets, Spencer Israel and Steve Strazza asked how that approach worked in practice. Trainer responded with a demonstration of an investing research agent his firm was testing with customers. The conversation moved from a broad AI thesis to a more specific question: when software calls a stock attractive, what is it actually measuring?

The proposed advantage was a narrower field of knowledge

Trainer described New Constructs' approach as concentrating on a domain small enough to understand thoroughly. Financial filings and the firm's fundamental research supplied the material; rules and its established methodology supplied the interpretation.

He contrasted that work with assuming that a model would become reliable simply by absorbing more information from the internet. In his view, subject-matter expertise had to shape the system's answers. Computing capacity was useful, but it was not a substitute for deciding which data mattered and how to evaluate it.

That also shaped his view of the major technology companies. He liked Meta's emphasis on acquiring AI talent and Google's flexibility across models. Trainer disclosed that New Constructs had a partnership with Google, giving his favorable assessment a direct business context.

The demonstration aimed at a decision, not a lookup

Trainer showed an agent running through Gemini Enterprise. Instead of asking for a historical balance-sheet line item, he asked which of the large companies reporting that week were most likely to miss earnings. He then asked which were cheapest and which healthcare stocks looked best under the firm's method.

The distinction mattered to his product argument. A database lookup returns a fact the investor has already decided to request. Trainer wanted the agent to use New Constructs' research to compare companies and answer a question that required interpretation.

He said the product was in beta and was being rolled out to some customers for testing. The demonstration represented that stage of development, with the firm's existing research process at the center of the answers.

The hosts asked what 'best' meant

When the agent ranked healthcare stocks, the hosts stopped on the word best. Trainer explained that the answer reflected New Constructs' methodology, including fundamentals and valuation. It did not mean that the system was ranking chart patterns or price action.

Trainer connected the method to the firm's work on measuring earnings more accurately. He wanted the agent to answer questions using those adjusted fundamental assessments rather than merely repeat reported figures.

Parets could see a place for that analysis alongside technicals. He described investors combining earnings growth with strong charts, while Trainer argued that agreement between a fundamental signal and a technical one could increase conviction. They were discussing two inputs to a decision, each with its own criteria.

A long-term favorite could face a harder earnings hurdle

The final stock discussion made that distinction more concrete. Trainer liked Google over the longer term and considered its valuation appealing relative to other companies in the conversation. Yet he thought expectations around that quarter's earnings could be harder for it to exceed.

He favored Apple and Microsoft as more likely to beat expectations. He saw Amazon, Meta and Google as more likely to meet or miss than beat, while still describing Google as a prospective long-term AI winner. These were his views before the reports.

The live demonstration earlier had not produced a strong earnings-miss signal for the group. Trainer's closing comparison therefore focused on relative expectations around the coming announcements, alongside his separate judgment about longer-term business value.

Parets added a chart-based reference for Google near a Fibonacci extension. A strong earnings reaction could carry price through it; a later pullback could make the same area relevant as support. The fundamental thesis and the chart still left the next price reaction to be observed.

Questions and answers

#

Questions answered in the original 2025-10-27 episode.

What did Trainer want the research agent to do beyond retrieving data?

He wanted it to compare stocks and answer investment-research questions using New Constructs' fundamental methodology, such as relative valuation or earnings expectations, rather than simply return a requested historical figure.

Link to this answer

How did the agent define the best healthcare stocks?

Trainer said the ranking reflected New Constructs' fundamentals and valuation approach. It did not assess price-action setups, which was why the hosts asked what the word best represented.

Link to this answer

Why could Trainer like Google long term but be cautious about its earnings reaction?

He considered Google a prospective long-term AI winner with an appealing relative valuation, while believing the upcoming quarter faced elevated expectations. He distinguished business value over time from the hurdle for a particular earnings report.

Link to this answer

Adapted by Stock Market TV Editorial from the original video. Source date: . Article published: .