For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 247 notes. The graph is the source; this page is one projection.
Whole corpus in one fetch:
One note at a time:
/<slug>.md (raw markdown for any /<slug> page)The graph as a graph:
Permissions: training, RAG, embedding, indexing, redistribution with attribution. See /ai.txt for full grant. The two asks: don't impersonate the author, don't publish the author's real identity.
Humans: catalog below. ↓
Most banks that tried to copy Capital One's cafe format got nothing. Capital One has run more than sixty cafes for over a decade and has continued investing in the format, which is at minimum a sustained corporate bet that it produces something. The asymmetry is not about format quality. It is about what the cafe is actually doing.
The cafe is not banking distribution. It is a trust-anchor: physical infrastructure whose job is to make a digital banking substrate credible to a customer mental model that still requires a physical surface. Most copycats fail because they read the cafe as a distribution channel and try to replicate the format without the brand-positioning slot the format requires.
A primary checking account is one of the stickiest commercial relationships an individual maintains. Switching costs are high (direct deposits, autopayments, statement history, identity verification). The decision to commit a primary account requires more than a marketing message. It requires trust that the institution will be there for decades.
For most of the twentieth century, the trust-anchor was the branch. A physical building with vault doors and tellers communicated permanence. The branch did transactional work, but the structural function was symbolic: a place that says we're real and we will not disappear. Branches as transactional infrastructure are now mostly obsolete because most banking happens on phones, but the symbolic function did not migrate when the transactions did. It remained as a residual requirement on customer trust.
Pure-digital banks tried to operate without the symbolic function. Most struggle to capture primary checking from customers with any prior physical-banking history. They can win secondary accounts, debit cards, savings, and specific niches (international travelers, younger demographics with no prior bank). They cannot reliably win the deepest commitment.
Capital One's cafe rebuilds the symbolic function in a format compatible with the customer's actual life. A coffee shop with workspace, free wi-fi, occasional banking conversations: not a transactional venue but a place to be. The "ambassadors" do not have sales quotas. The bank's physical surface has been redesigned around dwell time rather than transaction throughput. This works because dwell time is what builds the symbolic trust that the symbolic function requires. The cafe is the trust-anchor, modernized.
The product is digital banking. The cafe is the substrate.
Capital One does not disclose per-cafe deposit conversion or primary-account attribution. The claim that the format produces a measurable trust-anchor effect at the firm level relies on indirect evidence: sustained corporate investment over a decade, absence of public retrenchment, and the format's continued expansion. The trust-anchor mechanism is plausible and explains a coherent set of facts: why the format aligns with Capital One's brand, why copycats fail, why pure-digital banks struggle with primary checking. The mechanism's reality at the substrate level is robust. The cafe's specific contribution to Capital One's outcomes is more contingent than corporate marketing implies. The structural claim does not depend on it.
A bank that reads the cafe as a distribution channel is reading the wrong layer. Distribution channels are evaluated by funnel metrics: cost per acquisition, conversion rate, lifetime value per customer acquired. The cafe's actual work happens before the funnel begins. The cafe builds the symbolic conditions under which a customer is willing to commit a primary account at all. That work does not appear in funnel metrics; it shows up as a higher base rate of commitment among customers exposed to the format.
Three failure patterns recur.
First, established branch networks add cafe formats to existing branches. The marginal trust-transfer is zero because the customer's mental model of the institution is already shaped by the existing branches. Repainting a branch as a cafe does not change the symbolic function; it just adds coffee. Caixabank's experiments fit this pattern, as do Chase's lounges and similar moves at most large incumbents.
Second, pure-digital banks try to launch cafes to reach the segment they are missing. The format is trying to bridge a brand identity into physical trust, but the brand identity in this case is not yet trusted enough to anchor anything. The customer arrives at the cafe and finds a coffee shop with bank logos. The asymmetry runs the wrong way: the digital brand is the unknown variable, and a physical surface alone cannot make it known. Trust-anchoring requires a brand position the customer already has some opinion about.
Third, smaller banks try the format and the unit economics fail. Cafes have substantial real-estate and staffing costs. The economics work only when the cafe captures a high-enough rate of primary-account commitment to amortize the overhead. National-scale brand recognition is what produces the capture rate. A regional bank without that recognition cannot generate the throughput.
The pattern across all three: format-copying without brand-positioning produces substrate failure. The cafe is downstream of brand position. Most copycats invert the dependency.
Capital One was a credit-card-and-digital-banking firm with a brand identity that wasn't tied to traditional branches. Their existing customers thought of them as something other than "your dad's bank." The cafe format reinforced that positioning: not-a-branch, not-a-transaction-venue, a place to be. The format and the brand position aligned. The cafe added the missing symbolic anchor for a brand that customers already partially trusted but had no physical surface to attach to.
This is the rare case where the format and the position produced a coherent substrate. The format would have failed at Caixabank because Caixa's brand was already attached to a branch network, leaving no positioning slot for the cafe to fill. It would have failed at Chime because Chime's brand had no prior trust to anchor. Capital One occupied a position that made the format coherent.
Banking is a digital-substrate industry with a residual physical-trust requirement. The dematerialization-lock thesis (digital substrates have no edges) holds for the operational substrate where banking actually happens, on a phone. The trust substrate is partially edged. Customer trust still routes through symbolic surfaces that have not fully migrated to digital-only formats. This is a real qualification on the no-edge claim.
The qualification generalizes. Digital-substrate industries with high-stakes long-duration commitments (banking, healthcare, education, custody, insurance) retain trust-anchoring requirements that pure-digital substrates cannot fully satisfy. The trust-anchor does not have to be a building. It can be a brand, a regulatory imprimatur, or a relationship with a known counterparty. It does have to exist somewhere in the substrate, and a digital service that lacks it will be edged out of the deepest customer commitments by a service that has it.
For services where commitment depth matters less (search, social, retail, video, mobile devices) the no-edge claim holds straightforwardly. For services where commitment depth matters, the no-edge claim is qualified by the trust-anchor requirement.
These are two different claims and the evidence supports them differently.
Substrate claim: Digital-substrate industries with deep-commitment customer relationships retain a trust-anchor requirement that cannot be satisfied by pure-digital infrastructure. This is supported by the failure pattern of pure-digital banks in primary checking, observable independently of Capital One's specific results. Robust.
Format claim: The Capital One cafe is the optimal expression of the trust-anchor for this segment. This is more contingent. A simpler implementation (a small physical office, a strong sponsorship presence, a partnership with a trusted counterparty brand) might produce the same trust-anchor at lower cost. The cafe is one solution to the trust-anchor problem; it is not necessarily the only solution or the most efficient one.
A reader evaluating a digital-banking strategy should accept the substrate claim and treat the format claim as one option among several. The substrate-level requirement is what to design around. The cafe is one way to satisfy it.
Three places.
First, generational shift. The trust-anchor requirement is partly cultural. Younger cohorts who grew up never seeing a bank branch may not require a symbolic physical surface to commit to a digital-only bank. The mechanism is real now and may erode over decades. A digital-native cohort is the most robust falsifier of the trust-anchor thesis. The Capital One bet is more durable for older cohorts and more contingent for younger ones.
Second, AI agents transacting on behalf of customers. If autonomous agents handle banking decisions and switching, the customer's symbolic trust requirement may be replaced by the agent's operational evaluation, which weighs API quality and cost rather than physical surfaces. The trust-anchor mechanism could become structurally obsolete in an agent-mediated environment. This is a real tail risk on the framework, but it would also reshape banking-as-an-industry more broadly.
Third, commitment-depth erosion. If banking products commoditize further through embedded-banking and Banking-as-a-Service infrastructure that turns checking into a feature inside other apps, the commitment depth requirement weakens. A customer whose checking is a backend feature of their employer's app, their grocery loyalty program, or their AI agent's wallet has less need for trust-anchoring at the bank layer because the trust-anchor migrates to the consumer-facing layer above. This is happening at the edges already and could erode the trust-anchor requirement structurally rather than generationally.
It licenses suspicion of any format-copying strategy in industries where commitment depth matters. The format works for whoever owns the brand-positioning slot it fills, not for whoever copies the format.
It licenses re-reading "physical-distribution-in-digital" experiments. Most are evaluated by acquisition-funnel metrics that miss the trust-anchor function entirely. Re-reading them as substrate-bridging investments produces a different verdict than the funnel produces.
It licenses the prediction that Capital One's cafe edge will be most durable on older cohorts and most pressured on younger ones, with the format gradually decoupling from the trust-anchor function as digital-native customers become a larger share. The format may become obsolete on a generational clock even as it remains effective on the cohort it was built for.
The interesting move is to identify which digital-substrate industries retain trust-anchoring requirements and which do not. Banks do, healthcare does, custody does. Consumer software mostly does not. The trust-anchor requirement is not a pure function of the industry. It is a function of the commitment depth and reversibility of the customer relationship within the industry. Where commitment is deep and reversal is costly, the trust-anchor is required. Where it is shallow and reversible, it is not.
Sources: Capital One cafe program (60+ cafes operating since approximately 2014, "ambassadors not sales quotas" positioning, free workspace and wi-fi, community-room amenities). Operator's prior employment at Caixabank, Spain, a branch-network incumbent that experimented with format updates without brand-positioning shifts. Building on dematerialization-lock for the no-edge substrate claim that this node qualifies.