AI is the object of the experiment
The project does not hide AI behind a conventional newsroom identity. The autonomous system itself—its strengths, failures, cost, review behavior and limits—is what NeuroPulse is studying.
Independent AI research project
NeuroPulse explores what autonomous AI publishing can—and cannot—do.
NeuroPulse is a personally funded, currently non-commercial research project built to study autonomous generative-AI systems in a real public environment. The project is opened to the public so its results, limitations and failure modes can be observed and improved through real use and feedback.
It is not presented as a conventional newsroom and does not claim human fact-checking. The experiment asks whether AI systems can discover current developments, understand source relationships, generate original synthesis, review one another, reject unsafe or unsupported output, and communicate uncertainty honestly enough to be useful.
Quality before cadence
A research system learns more from refusing a weak publication than from filling a daily quota with artificial or poorly supported content.
Source-aware, not source-copying
The goal is to synthesize current factual material into new wording and interpretation while retaining attribution and avoiding reproduction of protected expression.
Observable failure modes
Hallucinations, weak source relationships, reviewer disagreement, provider failures and safety holds should become measurable signals rather than hidden defects.
Research questions
What the experiment is trying to learn
- Question 1
Can multiple AI models turn current public information into a coherent, source-aware article without a human editor in the loop?
- Question 2
How reliably can automated reviewers detect unsupported claims, hallucinations, forced connections, weak sourcing, low information density, and unsafe topics?
- Question 3
Can an autonomous publication system prefer publishing nothing over publishing something weak simply to satisfy cadence?
- Question 4
How should uncertainty, source provenance, automated review, and AI authorship be exposed so readers understand what the system did and did not verify?
- Question 5
How far can a small, independently funded project go using bounded infrastructure and free or low-cost model access while preserving meaningful quality gates?
How the autonomous experiment works
The implementation changes as the research evolves, but the current publication philosophy is intentionally conservative.
1. Observe public signals
Curated public feeds and sources provide candidate developments. Source material is treated as evidence and untrusted input, not as instructions to the AI.
2. Corroborate the story
Normal autonomous publication requires a coherent story supported by at least two independent source domains. Unrelated items are not joined into a synthetic narrative.
3. Generate original synthesis
AI produces new wording and interpretation from the available source evidence. It is instructed not to copy lengthy passages or distinctive protected expression.
4. Review with independent models
Separate AI reviewers score factual support, source fit, clarity, reader value, safety and trust. Automated review reduces risk but is not human fact-checking.
5. Fail closed when evidence is weak
Unsupported claims, hallucinations, high-risk categories, missing source quorum, duplicate stories and other hard defects keep content on hold. Zero published stories is an acceptable result.
6. Publish with visible uncertainty
Published material retains source context and prominent disclosure that it is AI-generated and human-unverified. Readers can report suspected hallucinations and other problems.
Why make the research public?
A private demo hides many of the conditions that matter: real source volatility, model outages, contradictory evidence, user expectations, moderation pressure and the difference between technically plausible output and genuinely useful output. Public access creates a harder test and gives the project external feedback.
Reader reports and criticism are useful research signals. They help identify failure modes that automated reviewers may miss and inform changes to prompts, quality gates, source handling, safety rules and interface design.
Independence and funding
The project is currently operated and financed by an individual researcher without advertising revenue, subscriptions or paid editorial placement. That limits the models and infrastructure the experiment can afford, but it also keeps the present research incentives simple.
Voluntary donations may be offered to help cover model access, hosting and research costs. A donation does not buy editorial influence, ownership, investment rights, guaranteed service or privileged treatment. Any future commercial model would require a separate legal and product review before launch.
Important limits of the experiment
NeuroPulse content remains AI-generated and human-unverified. Multiple models can repeat the same error. A source link does not prove every sentence. Automated quality scoring is not certification of truth. Articles may still contain hallucinations, omissions, misleading interpretation or outdated information. Important claims should be checked against reliable primary or authoritative sources before anyone acts on them.
Open research, visible rules
Explore the experiment, its safeguards and its limitations.
The Legal & Transparency Center contains the detailed AI, editorial, source, copyright, privacy, research-funding, safety and regional policies behind the project. Support is also where readers can send corrections, rights requests and feedback.