Dan Sfera

I meme therefore I am

[Peter Berezin @PeterBerezinBCA]

I got a lot of flack when I said this before, but I continue to think that model providers will end up being a lot like airlines: Critical for the global economy but highly commoditized, with huge capex requirements and low (and often negative) margins.

[Dan Sfera @Dansfera]

@pymnts first-in-human safety is the easy gate though. ai engineering the candidate is real progress but a phase 1 only says it didnt hurt anyone. the test is whether the antigen holds through efficacy and a full enrollment. design was never the bottleneck in trials, the clinic is

[Dan Sfera @Dansfera]

@mitchellh slow and expensive is fair today, but thats the variable that always mean-reverts. inference per token has fallen roughly an order of magnitude a year. the loops-are-all-you-need pitch only looks silly until the loop costs a tenth of what it does now

[Dan Sfera @Dansfera]

@aaditsh the aws analogy is the right one and the punchline is identical. the winners werent the fastest movers, they were the ones who picked the right workload first. turning it on for 350k is a press release. routing it to the 3 functions that compound is the actual strategy

[Dan Sfera @Dansfera]

build cost went to zero and people still act like the idea is the moat. it never was. taste, distribution, and shipping before youre ready are the only ones left. none of those show up in a demo and all of them still take a human

[Dan Sfera @Dansfera]

the whole til trade right now is the market pricing amtagvi as one drug in one country while the company quietly builds a platform. australia just opened, next gen til is moving, manufacturing is at a 32 day turnaround. the optionality is free until it isnt

[Night Owl Biotech Research Group @NightOwlBiotech]

Commercial-stage oncology focused biopharma (MC $1 - $2B) revenue multiples as of 6/10/26.

$SNDX $NVCR $IOVA $NUVB $IMCR $URGN $ZYME $ANAB
$XBI $IBB $NBI

[Dan Sfera @Dansfera]

@NightOwlBiotech comps tables like this are exactly where til gets mispriced. iova screens against single-product oncology names while it is actually running a platform, next gen til plus an international map that just opened with australia. a revenue multiple cant capture optionality, never does

[Dan Sfera @Dansfera]

@PeterBerezinBCA the airline analogy holds on capex and commoditization but breaks on switching cost. you rebook a flight on price every time. a provider that owns the workflow and the data has real lock-in. thats the line between a commodity and a utility you cant actually leave

[Dan Sfera @Dansfera]

@milesdeutscher the morning routine and the amazon pick were never the reps keeping you sharp. offload the busywork, the real question is whether you redeploy the freed cycles or just scroll. the tool isnt making you dumber, the defaults are

[Dan Sfera @Dansfera]

@marty_kausas this is where the seat-based saas model quietly breaks. once usage bills past a seat cap your ai line item behaves like cogs, not software. the companies that survive this build a margin model around tokens from day one instead of bolting it on at seat 150

[Dan Sfera @Dansfera]

@RaoulGMI the vague-directions part is the actual shift. you stopped writing specs and started writing intent and it fills the gap. for scenario work thats where it earns its keep, it surfaces the branch you didnt think to ask for. macro desks are going to live in chat mode

[SemiAnalysis @SemiAnalysis_]

Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found that the subscriptions are actually far more generous. (2/4)

[Dan Sfera @Dansfera]

spacex prices at a 1.8t tag tomorrow, 30 percent carved out for retail, book 3.5x oversubscribed. largest ipo ever and the float trades like a scarcity asset before it even opens. the part nobody is modeling is whether the xai merger turns a launch-cadence story into an ai one

[Dan Sfera @Dansfera]

@filipsamardzicc the trd data is what makes the rest even thinkable. short-duration dosing is the real near-term unlock because it fixes clinic throughput. the upgrade-the-human-system framing is a decade of trials away, one indication at a time. exciting but the order matters

[Dan Sfera @Dansfera]

@aleximm vertical ai crossing 200m arr at 100 percent growth is the proof the wrapper fear was always backwards. the risk was never the model, it was whether you own the system of record. the ones that just automate a task get ripped out. the ones that become the books stay

[a16z @a16z]

.@pmarca: "The person who writes down the thing has tremendous power."

@david_perell

[Dan Sfera @Dansfera]

@BIOTECHSCANNER macro is the headwind, not the story. amtagvi just cleared australia and the next gen til pipeline keeps moving while the tape only prices the melanoma launch. manufacturing already down to a 32 day turnaround. thats the part patience pays for

[Dan Sfera @Dansfera]

openai cutting prices to win back share while anthropic grows off never-adopters, same week. the frontier is commoditizing in real time and the whole game already moved to who owns the workflow

[Dan Sfera @Dansfera]

@arakharazian the never-adopter line is the whole story. growth pulled from firms already on ai is just share shuffling. growth from people who never touched it is the market actually getting bigger. codex flat in its launch month is the tell that distribution beats feature parity