House view
A house view is a stated conviction about structural direction, held separately from any single equity. We label conviction as conviction, and each one carries the evidence that would overturn it.
The industrialization of genetic cures
Casgevy is the first commercial proof that a single administration can functionally cure a genetic disease instead of managing it for life. If editing becomes standard of care across even a fraction of the roughly 7,000 known monogenic disorders, the modality shifts from niche therapy to the industrialization of genetic cures, and the addressable market is measured in the hundreds of billions.
The line between cure and enhancement is already blurring at the edges: CTX310 durably lowers ANGPTL3, LDL and triglycerides from a single dose, and delivered to a healthy person that same edit becomes prevention, not therapy. The credible near-term frontier is the disease-free human. Enhancement stays speculative for now. Heritable germline editing is banned across essentially every jurisdiction, and the population for genetic disease correction dwarfs any market for enhancement.
What gates it. Five structural constraints, two of them critical: delivery beyond the liver is unsolved, and the safety ceiling (off-target edits, chromosomal translocations, hepatic toxicity) remains the binding limit on how fast the modality can scale. Reagents and GMP manufacturing, reimbursement and treatment-center capacity, and foundational intellectual property make up the rest.
Amendment, 12 September 2026
As first published, this view said the safety ceiling sat under stricter FDA guidance requiring deep sequencing screens, and listed hepatic toxicity among the things that guidance covered. Working through the regulatory record for our Intellia initiation, we no longer hold that and are correcting it here.
The guidance exists: on 14 April 2026 the FDA issued draft guidance on safety assessment of genome editing using next-generation sequencing, recommending off-target analysis at sufficient depth and quantitative assessment of chromosomal translocations. But it is non-binding draft guidance covering nonclinical characterization, and it says nothing about hepatic toxicity. The liver monitoring now present in in vivo editing trials came from company protocol amendments negotiated to lift a clinical hold, not from the regulator. We attributed to the FDA something a sponsor did.
The second correction is directional. We described the regulator as tightening. Where it counts it is loosening: the Plausible Mechanism Framework of 23 February 2026 names genome editing explicitly and allows product variants to be evaluated under master protocols, and the Rare Disease Evidence Principles of September 2025 permit approval on a single adequate and well-controlled study in populations under roughly 1,000 patients. The FDA is raising the characterization bar and lowering the clinical evidence bar at the same time.
The safety ceiling stands as a constraint. It is set by observed clinical events and by what payers and physicians will accept, not by US regulatory posture. Read the full reasoning in Intellia, Section 08.
Why it is a national-security theme, not only a biotechnology vertical. States are treating capability in genetic medicine as strategic infrastructure. US intelligence has publicly stated that China pursued biologically enhanced military personnel; DARPA funds the defensive counterpart through its Safe Genes program; and the FDA now restricts routing American patient cells through China. Ownership of delivery, manufacturing and validation capacity is being treated by governments as a national asset.
- Expressed through bottleneck owners over sponsors, with CRSP the preferred sponsor
- Overturned by a serious safety event in a leading in vivo program, or by payers refusing single-administration economics at scale
Autonomous freight, and the insurance nobody is pricing
We believe autonomous technology in commercial trucking is not a question of whether but of when and by whom. The case is structural: a human driver is legally capped near eleven hours a day under hours-of-service rules while a driverless truck can run close to twenty-four, which roughly doubles the revenue a single asset produces. The driver is simultaneously the largest single line of operating cost and the largest single source of insured loss. When a technology lowers cost, raises asset utilization and reduces the worst losses at once, adoption becomes a question of engineering and policy timing, and demand takes care of itself.
The bottleneck is insurance, and it is usually treated as a footnote. Commercial auto liability is a $43B line running at roughly an 86% loss and expense ratio, the worst in five years. The apparent diversity of the market is an illusion: five parent balance sheets dominate it: Travelers through Northland, Old Republic through Great West Casualty, Berkshire Hathaway through National Indemnity, Markel through State National, and Chubb through ACE American. The top forty writers hold about 80% of the line. It is a hard market precisely because the insured event, a human making a mistake at speed in an eighty-thousand-pound vehicle, is so expensive when it goes wrong.
The regime change cuts three ways. Near term, premiums rise and the few carriers willing to underwrite a risk with no actuarial history gain pricing power. Long term, if driverless trucks are genuinely safer, the nuclear-verdict severity that crushed the line falls away and the legacy premium pool shrinks, while a new pool forms in product liability, technology errors and omissions, cyber, and fleet-level autonomous coverage, priced on data instead of driver records. The developers themselves may self-insure through captives and keep the liability pool inside their own economics entirely.
Amendment, 7 September 2026
As first published, this view held that whoever owns the telematics data prices the new risk best, and it named data-rich personal-lines incumbents as one end of the barbell. Working through Progressive's actual book for our initiation on that name, we no longer hold the second half of that claim, and we are correcting it here rather than leaving it to be discovered.
The data advantage is real. The entity was wrong. The new pool forms in large-fleet commercial and manufacturer product liability, which is written by the specialist and excess markets, not by a personal-lines generalist with a small-fleet trucking division. A carrier writing one-to-twenty-truck for-hire operators is standing in front of the disruption rather than positioned to underwrite what replaces it. The barbell now reads: specialist and excess writers and fronting platforms at one end, private insurtech and program specialists at the other, the pure trucking specialists a paired trade in the middle.
Read the reasoning in full in Progressive, Section 07. The allocation map in the Kodiak note predates this correction and is scheduled for restatement at its next revision.
- Expressed as a barbell, with specialist and excess writers and fronting platforms at one end, private insurtech and program specialists at the other, and the pure trucking specialists a paired trade in the middle
- Overturned by developers self-insuring through captives at scale, so the new pool never reaches public carriers, or by a high-profile driverless failure that resets the political timeline
Framework stances
Framework stances are internal analytical constructs, not regulated investment ratings. They describe the shape of the risk and reward we see at a stated horizon, and they are always paired with a risk table.
| Stance | What it means | Currently applied to |
|---|---|---|
| Constructive | The setup and the catalyst path point the same way at the stated horizon, with risks identified and survivable. Not a buy rating and not a target. | NTLA, CRSP |
| Speculative | A genuine option with a real prospect of total loss. Position sizing, not conviction, is the operative decision. | KDK |
| Neutral | The work is done and the asymmetry is not there at the current quotation. | PGR |
| Negative | The structure points against the equity, whether or not the underlying theme is sound. | Not currently applied |
Stances are reviewed when a note is revised. They are not maintained continuously and should not be read as live guidance.