US Jobs Miss, AI Twins, and Quant Research: Asset Growth Anomaly
Trending Topics
US September Non-Farm Payrolls Miss Expectations; Unemployment Rises to 4.2%
The private sector added 46,000 jobs, while the government sector shed 17,000. August and July data were revised down by a combined 60,000. Average hourly wages increased by only 0.1% month-over-month and 3% year-over-year, trailing the overall CPI increase of 3.4% in August. Healthcare added 17,000 jobs, below the 12-month average of 33,000. Following the report, equities rose as markets priced in no rate hike for October, while long-end yields may continue to climb. 1
EDGAR Tools Releases Final 5.x Version
edgar is an open-source market data tool for the SEC EDGAR system, capable of parsing filings such as 10-K, 13F, Form 4, and Schedule 13D/G, as well as XBRL and BDC portfolio data.
According to the update, the tool released v5.60.0 on 2026-10-02, marking the final 5.x version before 6.0. This release elevates holdings disclosure to a top-level import, adds a timeout parameter to the filing stream, and fixes 19 bugs, primarily related to duplicate calculations in BDC portfolios and the summation of joint filings.
Version 6.0 is scheduled for release on 2026-11-09 and will drop support for Python 3.10. 1
Industry News
Millennium Equips All Staff with AI Digital Twins
Millennium has rolled out its AI assistant, "Digital Twin," to all employees. The tool facilitates email collaboration and assists with research, meeting preparation, email management, and scheduled nighttime tasks. Initiated by CIO Vlad Torgovnik, the program now serves over 1,600 users, with 97% of early adopters using it daily. By default, the twins operate without privileged access, maintaining independent identities and audit trails. An "AI Teammate" feature for teams is planned for the future. 1
Citadel Securities Partners with Wolfe Research
Citadel Securities and Wolfe Research announced a collaboration combining the former’s liquidity and execution advantages with the latter’s top-tier research to create a trading execution service model for institutional clients. Citadel Securities also took a minority stake in Wolfe Research to ensure long-term synergy. 1
Point72 Executive Named to Top 50 Women in Hedge Funds
Point72 Managing Director Elena Bukowski was named to the 2026 Top 50 Leading Women in Hedge Funds list. Having been with the firm for nearly eight years, she has spearheaded institutional investor partnerships, driving AUM expansion. Point72 has been featured on the list for nine consecutive years. 1
Quant Research
The Root of the Asset Growth Anomaly: Mispricing, Not Risk
Alpha Architect Blog published new research challenging the risk-based explanation for the asset growth anomaly.
Traditional Fama-French and Hou-Xue-Zhang models attribute this anomaly to risk compensation. However, a July 2026 study by Zhuo Cheng et al. argues that the relationship between asset growth and returns is driven by mispricing. The study finds that companies with extreme asset growth or contraction exhibit U-shaped or inverted U-shaped mispricing distributions due to valuation difficulties and high arbitrage costs.
Through finite mixture normal regression analysis of over 172,000 observations from 1963 to 2022, the data supports a two-latitude structure: one group showing a U-shape (undervaluation correction) and another an inverted U-shape (overvaluation correction). This pattern remains significant after risk adjustment, and price-to-value ratios validate the mispricing hypothesis.
The study emphasizes that focusing solely on realized returns obscures this non-linear structure. The core of the asset growth anomaly lies in market mispricing rather than rational risk compensation. 1
Portfolio Strategy for Price Distortion: Simultaneously Robust to Trading and Learning
Learning from unobserved asset drift requires using observed prices, but the observation model used for learning may itself be misspecified. How to ensure that both trading and learning remain robust to perturbations in the price distribution remains an open challenge.
The study models a Bayesian investor who uses Kalman-Bucy filtering to estimate drift and trades accordingly. It introduces an adversary that distorts the observed price rule, penalized by relative entropy, such that the same distortion affects both wealth dynamics and drift estimation.
The analysis provides a characterization of the solution for path-space robust control, indicating that the optimal strategy requires endogenous correction of learning biases rather than treating trading and estimation separately. 3
Signal Feasibility Threshold: Evidence Accumulation Outpacing Value Decay
In live trading, signals require time to accumulate sufficient credibility from discovery, while the profit window shrinks simultaneously. Determining when investment is destined to be unrecoverable is a critical trade-off to address before deployment.
Methodologically, the approach converts accumulated evidence and remaining opportunity into the same Kullback-Leibler information scale for comparison. It derives closed-form feasibility conditions under a Gaussian specification and generalizes the information-theoretic lower bound to broader distributions.
The derivation characterizes the exact feasible information budget threshold and universal lower bound form, demonstrating that deployment is only viable when the evidence growth rate consistently exceeds the decay rate. This unifies capacity and lifespan into a single computable criterion. 7
Short-Dated Barrier Option Pricing Expansion and Dominant Term Characterization Under Diffusion Scales
When both the strike price and the knock-out level are tightly clustered around the spot price, providing a consistent approximation for extremely short-dated contracts in an environment of continuous volatility changes remains a pricing challenge.
The study assumes a continuous stochastic volatility framework, requiring the joint weak convergence and uniform integrability of the triple: normalized terminal log-return, volatility relative deviation, and interval maximum. The strike and barrier approach the spot price at a rate proportional to the square root of the time to maturity.
The proof demonstrates that the leading-order approximation in the limit can be given by a time-inhomogeneous Black-Scholes-type knock-out put formula adapted to this convergence mechanism. 9
Quant Career
Jane Street ASIC Reverse Puzzle and Solution Analysis
In August, Jane Street launched a unique hardware reverse-engineering challenge: the official release included only the physical GDS layout of a small chip, without netlists or internal signal names. Participants were required to deduce the chip's function and solve the puzzle.
The image above shows the puzzle on the left and the solution on the right.
The chip is a hardware validator for an 11×11 "Star Battle" (non-adjacent stars) grid puzzle. It serially reads the state of each cell over 121 clock cycles. It employs multiple 2-bit counters and ROMs to verify that each row, column, and specific region contains exactly two stars. It also uses delay lines to ensure no two stars are adjacent (including diagonally). Once all rules are satisfied, the circuit unscrambles the LFSR to output the通关 string (* TWO STARS *).
Participants' reverse-engineering approaches mainly covered the following directions: First, netlist extraction and simulation, using tools like KLayout and Magic to extract gate-level netlists via LVS, then converting them to Verilog or custom evaluators for simulation. Second, reverse derivation and dynamic probing, dividing modules based on physical "islands" in the layout to derive logic from back to front, or injecting pulse stimuli to observe register flips to locate constraints. Third, SAT formalization, unrolling the 121-cycle sequential circuit and using solvers to automatically reverse-solve for the winning input. Additionally, some participants chose to reverse the output end, bypassing grid rules to reverse-engineer the LFSR seed and ROM to restore the answer. 1