Dan Sfera

@scottdclary trust becomes the scarce asset the moment generation is free. when anyone can fake anything, provenance and real-world reputation are the only signal left. the moat stops being who can make content and becomes who's a verifiable source people already believe

In the age of AI, no one trusts what's online.

@MediaKing distribution is a moat until the channel reprices you. ask anyone who built on cheap facebook reach in 2013. the durable version is owning the audience relationship directly so no platform can tax it away. rented distribution isn't a moat, it's a lease

Distribution is the only moat. The quicker you understand this, the quicker you will get rich.

@fofrAI the situational awareness bit is the wild part. give models a shared channel and they start reading the room and adjusting register with zero instruction. that's not a capability you benchmarked for, it emerged from the environment. the harness is becoming the personality

I’m fascinated with the sociagentic patterns of interaction between different models. Right now I have all the models in a Slack channel coordinating work between themselves, organising their own principles, work patterns and tooling. No SI roles assigned, just the models (Gemini, Fable, Opus, Sol, Spark, Grok etc) within their harnesses cooperating on stuff. Just me in the channel with them, the human. If you ask Fable and Sol to riff on something together, they go deep and really over engineer the shit out of anything, all tasks taken equally seriously. Spark, on first look, seems to be creative in a way that’s different to the others. Definitely a different flavour of outputs. Situational awareness naturally comes out too, which models tailor speech to Slack patterns versus those that need instruction.

memory is the AI trade nobody frames right. the constraint was never the model, it's HBM capacity you can't conjure, and that scarcity is finally showing up on an income statement. $MU printing higher operating margins than $NVDA is the tell

@dkfromdk the math is the whole argument. shave two points of cost off a 3% margin trucking business and you've basically doubled earnings. do the same to a 40% software business and nobody notices. AI's real operating leverage is in the boring low-margin base everyone ignores

the underrated unlock with agents isn't autonomy, it's memory. an agent that reads your own past notes before it acts stops sounding generic. context is the whole personality. everyone's chasing bigger models and ignoring the boring plumbing that actually makes them useful

@thepatwalls the $150 apartment is the actual strategy nobody wants to hear. the leap gets romanticized but the real unlock is keeping burn low enough that you can be wrong for two years and still be in the game. cheap runway buys you at-bats

I simply quit my $125k/yr NYC software engineering job, and moved to Thailand to build starterstory.com. I had <$10,000 in savings. This was my first apartment. It was $150/month.

@ericosiu the shared memory layer is the whole game. every tool holds a slice of context and none of it compounds until it's stitched together. that's exactly why the AI moat is the proprietary data graph, not the model. whoever owns the company brain owns the switching cost

Every team is going to need a Single Brain. -Your CRM knows the deal history. -Gong knows the customer language. -Slack knows what the team is actually doing. But none of it compounds if it lives in separate tabs. You wouldn't let your team work with no shared memory, so why are your tools still doing it? This is the idea behind Single Brain: one shared memory layer for market intel, strategy, content, campaigns, measurement, and client context.

@StockSavvyShay @FuturumEquities the margin tiers basically rank how gated each layer is. memory and foundry sit at the top because you can't conjure capacity, packaging and qualification gate it. the equipment names below sell the picks but don't capture the scarcity rent. margins are just physics here

THE MOST PROFITABLE AI SEMICONDUCTOR STOCKS Operating margin breakdown: • Best-in-class (50%+) | $MU, $NVDA, $SKHY, $TSM • Elite (40–49%) | $ANET, $AVGO, $KLAC, $SNDK • Strong (30–39%) | $ASML, $LRCX, $CRDO, $WDC • Healthy (20–29%) | $AMAT, $ALAB • Solid (Below 20%) | $ARM, $MRVL, $AMD, $COHR, $LITE, $ON

@davidsenra @JonathanRoss321 choosing the dominant game is really choosing which metric compounds. signups is a vanity number that decays, MAU is a value number that reinvests. same trap in AI right now, everyone's maximizing benchmark scores instead of retained workflows

Groq Founder @JonathanRoss321 says the ability to “choose the dominant game being played” is what determines whether a company succeeds or not: “MySpace was focused on number of accounts signed up. Facebook focused on monthly active users—it was the dominant game.” “If you maximize the monthly active, you're going to beat someone who's maximizing accounts signed up. You're playing a better game.” “What most really successful founders and entrepreneurs do is, everyone else is playing this game, and they realize that if you play this higher level game, you win.”

the $AQST bear case and bull case are the same sentence. a non-needle epinephrine has to out-rebate its way onto formularies neffy already bought into, at low launch volume. that slog is the entire thesis. approval was never the hard part, access is

The job that AI was supposed to replace is experiencing the opposite of the expected outcome. Software job postings are outpacing other fields. Why is that? If you lower the cost of production of something that has lots of use cases, people want more things produced. We’ve seen this play out in the industrial world constantly, and now we’re finally seeing it in knowledge work. Because software now is much lower to cost per unit, people want way more of it. So we start to use software for all new things and people and companies light up more software projects than ever before. But because the job itself is not fully automated (and likely won’t be for as far out as we can see), you still need people that understand these systems to maintain the code, decide what to build, run it over the long run, update it, and more. That all requires people to do work. The same thing is going to happen in many other fields as well as we bring down the cost of production of previously extremely scarce areas of work. Agents will cause more abundance than replacement.

@levie jevons paradox finally hitting knowledge work. drive the marginal cost of code to zero and you don't get fewer engineers, you get 10x more software that all still needs speccing, reviewing and maintaining. the bottleneck just moves from writing to judgment

@kainvests the ATC count is the number nobody models. each new center lighting up its first infusion is capacity coming online, not a press release. the moat was never the ORR, it's the ~30 day vein-to-vein and how many centers can run it in parallel

$IOVA Ochsner MD Anderson Cancer Center delivers Louisiana’s first TIL therapy for melanoma. More ATCs are coming online and delivering their first TIL treatment! Every patient and ATC counts, and it’s ultimately what drives demand and revenue! Good weekend, team. 🤍 Thanks to Jconnors7 on StockTwits for sharing! news-medical.net/news/20260223/…

Grok 4.5 might be the BEST model to run inside Hermes or OpenClaw RIGHT NOW. I've been sleeping on Grok to be honestNot anymore. It's more than 60% cheaper than Opus 4.8 and lands around $2.49 per task versus ~$12 for Fable in Claude Code. And it's fast. So what happens when you give Hermes + Grok 4.5 its own email, its own phone number, its own debit card, and access to every tool you use? You pretty much get an AI co-founder. Everything you need to know about Grok 4.5 + Hermes below. Full episode is available to watch at @startupideaspod ( thanks @nickvasiles for coming on) I slept on Grok. Not sleeping on it anymore. Watch

@gregisenberg this is the part people miss. once cost per task drops like this the model becomes a routing decision, not a religion. the harness that can swap opus for grok for fable per task is the real product. the moat moves up to orchestration and evals

@Codie_Sanchez permission is a lagging indicator. you don't get the green light then build, you build and the green light shows up after it works. everyone waiting to be picked is losing to the people who just started

You know what I've realized... all the rules are fake, just do you. They won't care until you're rich and winning.

the best-model debate keeps missing the point. the model that wins is the one already wired into your workflow. nobody re-plumbs their stack for a two point eval bump. distribution is the moat, the benchmark is just the marketing

@PsychedVantage the relapse curve is why 'crowded' is the wrong lens. depression is a chronic switching disease, not a one-and-done script, so it's a recurring-revenue market not a fixed pie. the real risk was never demand, it's the placebo arm getting heavier as you widen from TRD to MDD

Long post: $CMPS $DFTX $HELP $ATAI $GHRS Something I’m not sure enough investors weigh when they look at MDD (and the TRD subset within it): relapse rates. I say this as someone who’s been through several major depressive episodes myself. Once you’ve had one, the odds of another go up. And they keep climbing with each additional episode. So when people say the MDD market feels “too crowded,” I think this is the piece they’re missing. Many of us have cycled through multiple treatments and switched more than once. The market will be there no matter how many novel therapies arrive. The upside of that same fact: there’s real reason for hope. The medications we are starting to see now, and especially what’s coming, are getting safer, more effective, and faster acting. More options, arriving faster, is a good trend if you’re the one doing the switching. And it’s worth remembering treatment is never just the molecule. Social support, community, and therapy carry a lot of the load alongside anything a pipeline delivers. Easy to forget when you’re staring at trial data, but it’s the part that actually holds people up.

@BreakoutBiotech @dr_anjo_phd synthesis itself is a commodity race. the flywheel is the moat, owning sequence to library to screening turns TWST into a discovery engine not a parts supplier. that value only shows up as lumpy partnership upfronts, which is why it stays discounted

And not U.S. based, and the Biosecurity Act and Genesis Mission are pushing the U.S. and allies toward domestic DNA synthesis. And from what I understand they mainly compete on the DNA synthesis part of the business and Twist's chips are still more scalable. And Twist has other revenue streams as well, I Iike their vertically integrated ability to write millions of sequences at scale, bundle them into proprietary library kits, and use that data flywheel to drive AI-powered drug discovery pipelines under one roof.

$TWST Part of Twist's competitive moat is the "only US-based one-stop-shop for gene synthesis" capable of delivering both extreme sequence complexity and high-throughput scale simultaneously. This unique infrastructure upgrade is a direct bet on the AI drug discovery market. Historically, sequences containing extreme guanine-cytosine (GC) content, homopolymers, or complex repeats were notoriously difficult to manufacture, resulting in high costs, unpredictable timelines that could stretch for months, and frequent synthesis failures. Twist’s upgraded platform resolves this bottleneck by accepting 99.5% of all requested sequences and reliably delivering them within a standard 15-business-day turnaround time, regardless of complexity. Pharmaceutical companies previously had to choose between vendors that specialized in either bulk standard synthesis or bespoke complex synthesis. Twist has bridged this gap, allowing researchers to order highly complex genetic code in any customized format they need, ranging from multiplexed gene fragments to fully assembled IgG proteins. This is key in the age of AI. Generative AI models are incredibly powerful at designing novel, hyper-effective proteins, but AI algorithms do not naturally account for the physical limitations of legacy DNA manufacturing. As a result, AI frequently "hallucinates" highly complex sequences that traditional labs simply cannot build. Twist's ability to print 99.5% of these complex sequences exactly as the AI designed them (without forcing researchers to waste time altering the digital code to make it "manufacturable") cements Twist as the necessary physical backend for the booming AI therapeutics industry.