How Lighthaus actually works

Where the intel comes from

Lighthaus is compiled by an automated research pipeline — not a team of analysts manually reading everything, but not a black-box hallucination engine either. Here is exactly what happens each week.

The pipeline sweeps a curated set of primary sources on a schedule: AI lab newsrooms (Anthropic, OpenAI, Google DeepMind, NVIDIA, Microsoft AI), model documentation pages, arXiv preprints, GitHub release notes, reputable tech news outlets (The Verge, CNBC Tech, MarkTechPost), and funding databases (Crunchbase, PRNewswire). Each source is fetched directly — not summarized at third-hand.

Each item in the brief must pass three gates before it goes in:

• A real, working source URL that you can click and verify yourself
• A factual claim that is supported by that source (no paraphrase that changes the meaning)
• A "so what" judgment that is clearly marked as editorial interpretation, separate from the reported fact

If a source cannot be verified or a claim cannot be grounded in a primary document, it is omitted. We prefer publishing eight accurate items over twelve uncertain ones.

Signal vs. launch-day marketing

The AI industry is unusually noisy. Labs publish benchmark numbers that are cherry-picked. Funding announcements describe markets that may not exist. "State of the art" is claimed by everyone simultaneously.

Lighthaus applies a simple filter: we report on things that change what a builder or operator should do this week. That means:

• Model releases with verified benchmark data from independent evaluators, not lab self-reports alone
• Pricing or access changes that affect API costs and product architecture decisions
• Funding rounds that signal category validation or competitive threat — with named investors and verified amounts
• Policy and regulatory developments that introduce operational risk for AI-dependent products
• Open-source releases with measurable adoption (GitHub stars, download counts) as a proxy for real-world utility

We do not report on demo videos without published weights or APIs, benchmark leaderboards without independent replication, or funding rounds with anonymous investors and unverified figures.

Preview vs. full weekly edition

The free preview on the main page shows the top four items from the current week's brief. The full brief, delivered every Monday morning, includes:

The brief is designed to be read in under ten minutes and to leave you with a specific short list of things worth investigating further. It is a starting point, not a final word.

What Lighthaus is not

Lighthaus is an informational research digest, not investment advice. Nothing in the brief should be construed as a recommendation to buy, sell, or hold any security, asset, or financial instrument. The "investing signal" framing describes venture capital activity and analyst thesis shifts — it is observation, not advice.

The pipeline can miss things. We do not monitor closed channels, paywalled primary research, or private company communications. Some developments break on social platforms (X/Twitter, LinkedIn) before they appear in indexable sources; those may appear in the following week's brief rather than the current one.

We are not affiliated with any AI lab, model provider, or VC firm mentioned in the brief. Mentions are based on newsworthiness and verified sourcing, not commercial relationships.