SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2604.00339 · 2026

When Career Data Runs Out: Structured Feature Engineering and Signal Limits for Founder Success Prediction

Yagiz Ihlamur

arXiv · PDF

Code that ran

5 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 failures of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2604.00339.json · badge: /api/ran/2604.00339.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
ihlamury/vcbench canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
apply_rules Ran
ran; our contract draft was wrong, not the code
ihlamury/vcbench/features/high_precision_rules.py
code served (permissive licence) · get_code("5d44bfc63b597737")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function apply_rules takes a single parameter: row
# Looking at the function body, it calls:
# - row.get('exit_count', 0)
# This means row should be a dict-like object with a 'get' method

# Create a test row that demonstrates both branches of the logic
row_with_exit = {
    'exit_count': 2,
    'name': 'test_company',
    'founded_year': 2015
}

ARGS = {
    'row': row_with_exit
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
best_threshold Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ihlamury/vcbench/evaluate.py
code served (permissive licence) · get_code("06c6f5f3e4dd4b87")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create synthetic ground truth labels (binary classification)
y_true = [0, 1, 1, 0, 1, 0, 1, 1, 0, 0]

# Create synthetic predicted probabilities for class 1
y_prob = [0.1, 0.9, 0.8, 0.2, 0.85, 0.15, 0.75, 0.95, 0.3, 0.25]

ARGS = {
    'y_true': y_true,
    'y_prob': y_prob
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
evaluate Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ihlamury/vcbench/evaluate.py
code served (permissive licence) · get_code("db8ef4f5f3b7a462")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create small sample data for binary classification evaluation
y_true = [0, 1, 1, 0, 1, 1, 0, 0]
y_prob = [0.1, 0.9, 0.8, 0.2, 0.7, 0.95, 0.3, 0.15]
threshold = 0.5

ARGS = {
    'y_true': y_true,
    'y_prob': y_prob,
    'threshold': threshold
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
predict_one Ran
ran, honoured the contract we drafted
ihlamury/vcbench/baselines/zero_shot_baseline.py
code served (permissive licence) · get_code("1e3e3e33602aee8c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import json
from unittest.mock import Mock, MagicMock

# Create a mock client that simulates the Anthropic API
mock_client = Mock()

# Create a mock response object
mock_response = Mock()
mock_message = Mock()
mock_message.text = json.dumps({"probability": 0.75, "reasoning": "Strong technical background with prior exits."})
mock_response.content = [mock_message]

# Configure the mock client to return our mock response
mock_client.messages.create.return_value = mock_response

# Example prose input for a founder profile
prose_input = "Founder John has 10 years of software engineering experience at FAANG companies, previously founded a successful B2B SaaS startup that was acquired for $50M, and is now starting an AI infrastructure company with a strong technical co-founder team."

ARGS = {
    'client': mock_client,
    'prose': prose_input
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sweep_thresholds Ran
ran; our contract draft was wrong, not the code
ihlamury/vcbench/evaluate.py
code served (permissive licence) · get_code("00a0e42cd2660f59")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create small synthetic data for binary classification
n_samples = 20

# Ground truth labels (0 or 1)
y_true = [0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1]

# Predicted probabilities for class 1 (values between 0 and 1)
y_prob = [0.2, 0.8, 0.75, 0.3, 0.9, 0.4, 0.1, 0.7, 0.85, 0.25, 
          0.6, 0.95, 0.15, 0.35, 0.88, 0.92, 0.22, 0.78, 0.3, 0.87]

ARGS = {
    'y_true': y_true,
    'y_prob': y_prob,
    'steps': 50
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
add_llm_features Not yet run
unverified (harvested, no recorded run)
ihlamury/vcbench/features/extract_llm.py
code served (permissive licence) · get_code("7f82a41de50e5af8")
How it ran

no invocation recorded.

extract_features Not yet run
unverified (harvested, no recorded run)
ihlamury/vcbench/features/extract_structured.py
code served (permissive licence) · get_code("cbc1c01f13a27aff")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 7 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Predicting startup success from founder career data is hard. The signal is weak, the labels are rare (9%), and most founders who succeed look almost identical to those who fail. We engineer 28 structured features directly from raw JSON fieldsjobs, education, exits-and combine them with a deterministic rule layer and XGBoost boosted stumps. Our model achieves Val F0.5 = 0.3030, Precision = 0.3333, Recall = 0.2222-a +17.7pp improvement over the zero-shot LLM baseline. We then run a controlled experiment: extract 9 features from the prose field using Claude Haiku, at 67% and 100% dataset coverage. LLM features capture 26.4% of model importance but add zero CV signal (delta = -0.05pp). The reason is structural: anonymised_prose is generated from the same JSON fields we parse directly-it is a lossy re-encoding, not a richer source. The ceiling (CV ≈ 0.25, Val ≈ 0.30) reflects the information content of this dataset, not a modeling limitation. In characterizing where the signal runs out and why, this work functions as a benchmark diagnostic-one that points directly to what a richer dataset would need to include.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2604.00339")
get_code_for_paper("2604.00339")
have("2604.00339")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2604.00339.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2604.00339.svg)](https://syntology.ai/paper/2604.00339)

Connect an agent — have() is free.