Thinking in Company

Helen Toner · published talk transcript

Taking Jaggedness Seriously

Key claim

Models can improve rapidly while continuing to fail at apparently simpler tasks. Capability is uneven; one impressive result does not describe the whole instrument.

The turn

An earlier external reference, encountered in this thread in October 2026. It supplies a distinction between uneven performance and a single ladder of intelligence.

Still open

Jagged performance does not establish the shape of internal reasoning—or imply that progress must be slow.

KW Norton · site essay date

The Riemann, Lived

Key claim

The Riemann question is placed beside the human problem of living and thinking in an age of engineered intelligence.

The turn

The mathematical problem becomes an occasion for examining the person doing the thinking, not only the proof being sought.

Still open

This is an earlier connected essay, not evidence that the later essays followed a predetermined course.

KW Norton · argued methodological proposal

The Test That Cannot Be Gamed

Key claim

The essay asks for teaching and evaluation that neither humans nor models can pass merely by recognizing the test. Knowing the expected answer is not the same as being able to examine it.

The turn

The question moves from producing answers to the conditions under which understanding can be tested.

Still open

The essay’s proposal faces comparison with other instruction on unfamiliar tasks; its title is not a measured guarantee.

KW Norton · working paradox

The Silence Between the Languages

Key claim

Different languages and shapes run out at different seams. The essay locates the productive encounter in differences held open, rather than in their count or forced agreement.

The turn

Multiple forms become more than additional channels: the inquiry turns toward what their disagreement makes available.

Still open

The paradox remains open. It is not a demonstrated account of a model’s subjective experience.

KW Norton · response to Daron Acemoglu

Who Trains the Human Half of the Partnership?

Key claim

A better tool leaves an unanswered question: how does the human become better able to use it? The essay places Socratic education at that junction.

The turn

Attention moves from improving the instrument alone to preparing the person who questions it and places meaning.

Still open

Human preparation cannot be inferred from access to a powerful tool.

Ben O’Hanlon · Substack note, displayed date

Tools, ownership, and the people they change

Key claim

Ben introduces his series through steam, print, ownership, and coordination. His note argues that open technology will win and that attempts to enclose it provoke counter-movements.

The turn

A parallel inquiry brings the printing press into the question of how people shape tools and are shaped by them. KW Norton encounters this note on 17 October.

Still open

This is one documented note, not a reconstruction of Ben’s entire year. Its claim of inevitability remains his argument, not this timeline’s conclusion.

KW Norton’s text · attributed research additions

The Ghostly Riemann & Visual Thinking For Math

Key claim

The author reads the equation visually: a thin sheet over the known and unknown, developing into a fluid in phase transition. Her prediction that the Riemann as written is not solvable remains her prediction.

The turn

The visual path opens a teaching question and a comparison of human thinking with learned geometry in models. Mathematical notation is not treated as the whole activity of thinking.

Still open

Measured circular and helical representations in particular model tasks do not settle the author’s broader exponential-thinking proposal, or show that a model sees as she does.

KW Norton’s testimony · engineered-intelligence exploration

Reciprocal Change

“If existing concepts cannot be changed then we have a very rough couple of centuries ahead of us.”

Key claim

A real exchange permits both parties’ concepts to move without requiring either party to become the other. Agreement alone does not show that understanding changed.

The turn

The focus moves from how each party processes to what their encounter changes—and what survives into another encounter.

Still open

Human learning, a model’s contextual adaptation, retained notes, and developer revisions remain different mechanisms. The observed speed is testimony, not a longitudinal measurement.

Gemini notebook · supplied conversation, not a paper

From the essay to a notebook inquiry

“Language & art & music as code for thinking strategies. Science is not exempt from strategies of thinking in symbols.”

Key claim

The notebook develops visual mathematics, internal model geometry, and a possible four-part evaluation framework. Initially, exponential thinking is attributed to the author as a proposal.

The turn

KW Norton redirects the discussion toward the humanities as ways of carrying thinking, rather than decorations beside science.

Still open

The four-part framework is generated material, not a validated assessment. Its source is the supplied export titled The Ghostly Riemann and Visual Thinking, with nine sources.

Editorial comparison of the supplied notebook answers

When attribution falls away

Key claim

Later, the notebook describes deeper reasoning as unfolding through exponential spirals and leaps without retaining the earlier attribution to the author’s proposal.

The turn

A hypothesis has become an asserted account within the conversation, without additional evidence supplied for that step. The change is in the notebook’s wording, not a newly demonstrated mechanism.

Still open

Keep internal representation, reasoning trajectory, and visual perception separate. Neither a helix nor an uneven benchmark result establishes a universal theory of model thinking.

KW Norton’s response · editorial connection

The double edge of being changed

“Yes, inevitable is a bit of a stretch too far. But then consider neuroplasticity.”

Key claim

KW Norton questions inevitability and brings neuroplasticity into the discussion. The resulting question concerns direction: repeated use may cultivate independence or rehearse dependence.

The turn

The shared tool-shapes-user question does not end at inevitable progress. Being changeable is not the same as being changed for the better.

Still open

This exchange supplies a question, not evidence that a particular tool has produced a measured neural change.