The Quant Trading Module — An Honest Statement

This article is different from every other in the series. It's not "look at this great architecture." It's "there was supposed to be an architecture here, and there isn't." Plainly: the current repository contains no quantitative trading implementation code. Everything you see labeled internal/ares_quant, internal/quant, and examples/quant-trading in the old version of this article, in CAPABILITY-MAP, in ARCHITECTURE, and even in a dedicated docs/en/development/quant-trading.md — does not exist.

一、The honest conclusion, up front

I grepped the entire repository. The result is unambiguous:

If you grep quant, the hits are misleadingly unrelated:

• internal/fabric/task/quantum.go and internal/kernel/quantum_hook.go — these are execution quanta (a "one execution step" concept in DAG orchestration), nothing to do with trading.
• The word "quanta" in docs/25-config-yaml-guide is the same scheduling concept.
• grep position / trading / strategy hits resolve to regex match positions, strategy_adapter.go (an evolution strategy), and progress — all unrelated.

The old article claimed the quant module had 9,768 lines (~11% of the codebase), a market-making engine, portfolio risk metrics, a backtesting framework, research agents, Yahoo/CoinGecko/Polymarket data sources, SQLite storage, MCP tool registration… None of that is backed by any code. Those claims should be flagged unverified（待核实） — but the truth is stronger than "unverified": it's "verified absent."

二、So what about "quant" actually survives in the repo?

Not much — but one thing is real: a document, not code.

2.1 docs/en/development/quant-trading.md (removed 2026-09-14)

This was a design guide / implementation plan — literally titled "ares for Quant — 量化交易开发指南." It describes what should be built: 8 agent roles (fundamentals / sentiment / news / technical analysts, bull & bear researchers, trader, risk manager, portfolio manager), plus an architecture for internal/quant/ (~1,850 lines) and examples/quant-trading/ (~1,710 lines) with line-count estimates.

But the giveaway is that it's a blueprint, not a snapshot of reality: the interfaces it cites are all in deleted or nonexistent packages.

The doc also references plan/quan/quant-implementation-plan.md — that plan file does not exist either. The best this document can serve as evidence of is: "we planned it, even wrote a blueprint, but either the code was never landed or it has since been removed."

2.2 The "unverified（待核实）" entries in CAPABILITY-MAP and ARCHITECTURE

• docs/reference/CAPABILITY-MAP.md / docs/reference/CAPABILITY-MAP.en.md:

  Quantitative trading | internal/ares_quant | Market making, indicators, portfolio management, research

  A row that lists internal/ares_quant, yet no such package exists in any code directory. (待核实: it may describe an removed or never-merged version.)

• docs/zh/ARCHITECTURE.md (removed 2026-09-14, superseded by the repo-root ARCHITECTURE.md):

  • draws QUANT["internal/ares_quant<br/>Portfolio / Market / Research / MarketMaking / Indicators"] in its diagram
  • tabulates internal/ares_quant | 量化 | 投资组合模拟、市场数据、研究记忆

  These diagrams and tables describe modules on paper, not implemented modules. There is no ares_quant in internal/.

2.3 The old article itself (real, but its content is not)

docs/articles/zh/23-quant-trading.md and docs/articles/en/23-quant-trading.md exist, but their content is fabricated — they describe a module that doesn't exist. This rewrite (and its Chinese counterpart) exists precisely to correct that.

三、Framework capabilities that genuinely exist (what a quant build could rest on)

To avoid misleading people, let me be precise about which ares capabilities cited by the quant blueprint are real:

In other words: nothing stops you from building a quant research system on ares — the framework's capabilities are all here — but the ares repo itself ships no trading logic. To use it you'd build those modules from scratch yourself, not import a ready-made internal/ares_quant.

四、What this says (a lesson, not an excuse)

I came to write "here's the quant module we built." To write it truthfully I actually went and read the code — and found there was none. That alone is a loud wake-up call, worth writing down:

1. Docs and reality drift apart — faster than you'd think. CAPABILITY-MAP, ARCHITECTURE, and the old article still talk about a module that was removed or never landed. When the paths those docs cite (internal/dashboard) are also gone, the drift compounds.
2. The flip side of "experiments should be labeled as experiments" is this: a module that doesn't exist shouldn't masquerade in the docs as if it does. Blueprints and current state must be labeled separately. This rewrite strikes the "fake-existent" module off the capability list, restoring it to an honest "planned/removed" state.
3. The framework's real value is what deserves the spotlight. ares' solid core is Runtime + Workflow + Memory + Events + Flight Recorder. Letting a "quant demo" impersonate a core capability only makes people who try to use it come up empty.

The most honest closing line: the current (0.3.x) ares repository has no quantitative trading module. If you see internal/ares_quant written anywhere, it's either stale documentation or an as-yet-unlanded blueprint. The codebase has no such thing.

Appendix: verify it yourself

cd /Users/scc/go/src/goagent
find . -type d -name "*quant*"                              # only quantum.go scheduling code, no ares_quant
grep -rli "quant\|marketmaking" internal/ examples/        # hits are quantization/quantum/unrelated
grep -ri "ticker\|sharpe\|drawdown\|backtest\|polymarket\|coingecko" internal/ examples/  # no trading code
ls examples/   # no quant-trading

Every other architecture article in this series describes code that actually exists; this one (article 23) is the series' honest-disclaimer exception.