← Hacker News

Show HN: TopoGlyph – an extensible language system for problem solving

Hacker News · zwyld · June 15, 2026
TopoGlyph is an extensible language system designed to enhance language model problem-solving by enabling the generation of novel symbols that visualize problem structures. The system is available as a Claude Code plugin on GitHub and has been recorded on the Ethereum blockchain for preservation. The approach was inspired by concepts from the APL programming language.

Detailed Analysis

TopoGlyph is a user-created symbolic language system designed to augment the reasoning capabilities of large language models, particularly Claude, by providing a structured framework of glyphs and notation for modeling and solving complex problems. The project, available on GitHub as a Claude Code plugin, operates on the premise that prompting an LLM to first extend TopoGlyph's vocabulary for a given domain — inventing new symbols as needed — and then apply that extended notation to a problem produces measurably more structured and analytical outputs than conventional prompting. The creator acknowledges that the core module contains "hallucinated" concepts and conclusions, framing this not as a flaw to be resolved but as an emergent and interesting property of the system: the models invent internally consistent symbolic scaffolding even when the underlying conceptual grounding is imprecise.

The intellectual lineage the creator invokes is significant. The explicit comparison to APL — Kenneth Iverson's terse, symbol-dense array programming language from the 1960s — signals a particular philosophy: that the right notation can fundamentally change what kinds of thoughts are expressible and how efficiently reasoning can proceed. APL's core insight was that compressing complex operations into single glyphs reduced cognitive overhead and revealed structural patterns otherwise obscured by verbose syntax. TopoGlyph applies this same logic to natural language reasoning rather than computation, essentially asking whether a purpose-built symbolic vocabulary can serve as cognitive scaffolding for an LLM working through a hard problem.

The creator's decision to inscribe the full TopoGlyph text onto the Ethereum blockchain reflects a notable anxiety about the fragility of AI-adjacent tools under shifting corporate policy. The reference to Anthropic's "Fable model" being deprecated — the creator's uncertainty about the precise term ("banned") is itself telling — points to a broader concern among third-party developers building on top of proprietary AI systems: that models, capabilities, or access arrangements can be withdrawn with limited notice, stranding dependent tooling. By anchoring the specification to an immutable on-chain record, the creator is hedging against the possibility that TopoGlyph could become inaccessible if Anthropic were to restrict or alter the Claude ecosystem in ways that break the plugin.

This project sits at an active frontier in AI research and practice: the study of how prompt structure, meta-languages, and symbolic scaffolding affect the quality of LLM reasoning. There is a growing body of work — ranging from chain-of-thought prompting and scratchpad methods to more formal notational systems — exploring whether giving models richer intermediate representational tools improves their outputs on complex tasks. TopoGlyph is an informal, community-driven contribution to this space, distinguished by its extensibility and its embrace of model-generated symbolic invention as a feature rather than a risk. The creator's observation that the system "works anyway" despite containing hallucinated concepts raises a genuinely interesting question about whether structural coherence in a notation system can produce useful reasoning even when the semantic grounding of individual elements is loose — a question with implications for how AI systems handle novel or underspecified domains.

Read original article →