Independent research · self-funded · held to account

Understand the companies behind the AI buildout.

Dragonfly Lens turns filings, policy changes and industry data into plain-English research on AI infrastructure, energy and critical materials — with the sources, the risks, and a corrections log we never delete.

Read this week's research → See the track record
Updated weekly

The AI build-out scoreboard

Four checkable tests, marked with dates and sources. The “vs last week” column is the point: most weeks several cells say “no new print” — the data behind them is monthly or annual, and we would rather show that than refresh a stale number.

Test Latest vs last week Direction
Savings or credit?Debt ~$35.6bn; newest paper junk and priced widernew printWEAKENING
Capacity ahead of demand?H100 rents rose, not fellsameSTRENGTHENING
Circular financing?A guarantee of up to $105bn, signed in AugustsameWEAKENING
Is the narrative doing the work?The loudest argument in the marketsameLIVE — not graded

See the full scoreboard, with the five physical tells and every source →


How we work

Built to be checked, not just believed.

Three rules. They are the whole method, and you can hold us to all three.

Source-linked claims

Every claim is sourced, and fact is kept separate from hype. We route you to the primary source document — the filing, the grant, the rule change — not a summary of someone else's summary.

Risks, and what would prove us wrong

Every piece states the specific conditions that would break the thesis — what would have to be true for us to be wrong. Positions are held until the thesis breaks; we do not publish sell rules.

Public corrections

When we get something wrong, we say so — publicly, dated, and never deleted. The misses sit on the shelf next to the wins, on purpose. See the corrections log →

One thing to be clear about: our historical results are a labelled backtest — what the methodology would have flagged across public data, not money that was traded. The forward, dated record began in May 2026 and is still young. Judge us on what comes next.
See the track record →  ·  See the corrections log →

Backtested, losses included
87 dated theses. Median +21%/yr against +15%/yr for the S&P 500 over the same windows — every position still open, so those returns are unrealised.
The method, the losers, and how the +21%/yr is calculated →
Forward record
Dated and signed from 17 May 2026. So far: 1 signed signal, 0 settled outcomes. It is small, and we would rather say the number than round it up.
See the live count →
When we are wrong
Corrections are dated, kept beside the original claim, and never deleted — including the time we overstated our own track record and published the arithmetic.
Read the corrections log →
Pricing

Three tiers. No lock-in, cancel any time.

The research above is free. A subscription adds alerts, tooling and depth on top of it. First 30 days, full refund — no questions asked.

Basic
$9.99 / month
or $89.99/yr — save $30/yr
For getting familiar with the system and the sectors we track.
Next week you get: the Friday digest, plus the week's research in full.
  • Weekly digest every Friday
  • 3-ticker watchlist
Get started →
Premium
$36.99 / month
or $359.99/yr — save $84/yr
For investors who want the full context behind every signal.
Next week you get: the above, plus the two pages only this tier gets — the scoreboard’s working copy and the invalidation sheet — republished the same Friday as the free scoreboard. Radar and Compound come with the tier; they are not the reason it costs more.
  • Everything in Investor
  • The scoreboard’s working copy — what each cell means for the chain, with the primary document behind every figure
  • The invalidation sheet — the print that would break each thesis we have published, and when we last checked
  • Research notes on published signals
  • Radar — 4-tier coverage of every ticker we track
  • Compound — the trade journal
  • Unlimited watchlist
Subscribe →

We never sell your data — and we never even see your card. By subscribing you agree to our Terms & Risk Disclosure.

About

Built by someone who actually trades it.

I spent most of my life doing carpentry, and over 15 years figuring this out. I am not a hedge fund manager and not a financial advisor — this is editorial research. I built Dragonfly Systems because I got tired of tracking all of this by hand; the system does it now. I publish it because more people should have access to this kind of research, and because keeping a public track record keeps me honest. The name comes from the dragonfly — reportedly the most effective hunter in the natural world. Not brute force: precision. Not more signals, better ones.

Questions or feedback? Write to officialdragonflysystems@gmail.com. I read everything.