Editorial Standards
Our editorial process is not a checklist. It is a system designed to catch what others miss. Here is exactly how we work.
Our non-negotiables
Five rules shape every story we publish. Break one, and the piece gets rewritten or killed.
Primary sources only
We consult to the company who built the model, wrote the paper, or shipped the product. PR statements are starting points, not endpoints. If we cannot trace a claim to the authentic source, we do not publish it.
Verify the benchmarks
Every performance claim gets checked against the original benchmark data. We look for cherry-picked metrics, hidden evaluation conditions, and comparisons that conveniently omit stronger competitors. Good numbers are not enough. Honest numbers are.
Context over speed
We do not race to publish first. We race to publish right. A story that lands two hours later but includes the full technical context, competing viewpoints, and historical precedent beats a rushed headline every time.
Disclose the limitations
No model is perfect. No tool works everywhere. We report the failure modes, the edge cases, and the scenarios where the technology falls apart. Our readers are smart enough to handle the full picture.
Correct quickly
Mistakes happen. When they do, we fix them within 24 hours and mark the correction clearly at the top of the story. We do not bury updates at the bottom. We do not pretend errors never happened.
From idea to publish
Every story moves through four stages. Each one has a clear purpose and a clear gatekeeper.
Source
Identify the primary source. Company newsroom, Paper, commit, or direct conversation. No third-hand claims.
Scrutiny
Stress-test every claim. Look for contradictions, omissions, and conflicts of interest.
Screen
Editorial review against our five standards. Rewrite or kill if any rule breaks.
Publish
Final fact-check, copy edit, and live monitoring for corrections post-publish.
How we use AI in our workflow
We do not pretend AI does not make our work faster. It does. But we use it as a tool. Hers’s how AI helps us:
- AI-assisted research scanning across 60+ sources daily
- Automated extraction and cross-referencing
- Semantic analysis of press releases & news against historical claims
- Pattern detection in funding rounds and talent movements
- First-draft generation for routine announcements
- Rigorous editing and fact-checking before publishing
Corrections & Updates
We correct errors within 24 hours of discovery. Every correction gets a clear note at the top of the original story, including the date of the change and a brief explanation of what was fixed.
If you spot an error — factual, contextual, or technical — send us a note . We read every message and respond to corrections within one business day. We do not hide behind anonymity. We stand by what we publish, and we fix what we get wrong.