Felix T.

They cite the jam study, where shelves with 24 options produced far fewer purchases than shelves with 6, to argue that the industry scaled decision fatigue into a huge market. E-commerce conversion rates still sit at a few percent despite thirty years of optimisation, they note, and the goal should be to remove the need to browse rather than build a better search box.

One post opens with the author trying to buy protein powder on Amazon, seeing 1,847 results, closing the tab and buying nothing.

• It contrasts that with a 1995 supermarket carrying around 7,000 products and Amazon carrying 350 million, then returns to the jam study and Barry Schwartz's book on choice to explain why more options made the purchase less likely.
• Another post makes the same case with Prime Day 2025 data: AI-driven traffic to US retail sites rose 3,300% year over year but was still under 9% of total traffic, leaving most shoppers searching and scrolling as before.

The argument comes in two forms: a personal shopping failure and a market-data observation about how little retail behaviour has changed.

Read original · Original 2

They cite named launches and standards: Mastercard and Google's Verifiable Intent trust layer for proving a human authorised an agent purchase, Google's Universal Commerce Protocol co-developed with major retailers, Stripe and OpenAI's Agentic Commerce Protocol running live transactions, and a reported 15x rise in AI-powered Shopify orders since January 2025. The remaining gap, they argue, is perceptual: infrastructure has advanced faster than most people's understanding of it.

A post recalls telling friends in 2017 that they were going all-in on AI, when the word 'agent' still meant an insurance salesman, and compares the feeling to the internet in 1996. Eight years later, the post lists Verifiable Intent, the Universal Commerce Protocol, the Agentic Commerce Protocol, $115-135 billion of infrastructure spend tied to Mark Zuckerberg's stated priority, and a Morgan Stanley projection that half of US online shoppers will use AI agents by 2030.

A companion post sharpens the point: the protocols are real, but the commerce surface is not yet wired up. Claims of readiness, they argue, should be seen as part of an active debate about adoption, not a settled fact.

Read original

Payments, identity and trust are often listed as blockers, it argues, even though x402, Stripe and PayPal agent SDKs, Google's AP2, MCP with OAuth 2.1, and DIDs inside EU eIDAS 2.0 wallets already exist.

• The remaining bottlenecks are cross-ecosystem trust: an agent on one model provider cannot easily act for someone whose other agent runs elsewhere, and no neutral registry vouches for an agent's identity or spending limits; dispute resolution: no legal framework decides whether a bad agent purchase is charged back to the merchant, the model provider or the agent runtime; and adoption: MCP server use is growing internally, but almost none of it is customer-facing.
• The post describes Rezolve's work as the adoption surface, not a new set of protocols.

The post describes an agent running on Anthropic trying to act for someone whose other agent runs on OpenAI or a fine-tuned Llama in a private VPC, with no neutral registry to confirm it represents a named person. It also argues that consumer APIs such as Airbnb, Uber, Shopify and Booking need to wire up MCP or A2A with payment-ready mandates before agent-buys-for-you demos stop being theatre.

Read original

The intended experience is one sentence from the buyer, maximum context extracted, and the right product served in a conversation rather than a catalogue dump.

One post recounts firing a Dragunov, a Desert Eagle and a Tavor at Battlefield Vegas.

• The instructor did not empty every weapon onto the table and tell the author to pick one.
• Instead, he watched the author's grip and stance on the first gun, then chose the next based on what he saw—no menu, no filters.
• The post contrasts that with online shopping, where buyers face 4,000 products, sort by price, filter by colour and page seven, with nobody watching or adapting to what they actually need.

Read original

It is the only post here with a personal timeline, so newcomers should open it first.

Felix Tseitlin recalls telling friends in 2017 that they were going all-in on AI, when the word "agent" still meant an insurance salesman, and compares the feeling to the internet in 1996.

• The post lists what has recently arrived: Mastercard and Google's Verifiable Intent trust layer for proving a human authorised an agent to buy; Google's Universal Commerce Protocol, co-developed with Shopify, Walmart, Target, Etsy and Wayfair; Stripe and OpenAI's Agentic Commerce Protocol running live transactions; $115-135 billion of infrastructure spend tied to Zuckerberg's stated priority; a 15x rise in AI-powered Shopify orders since January 2025; and a Morgan Stanley projection that half of US online shoppers will use AI agents by 2030.
• The closing argument is that the gap between where the infrastructure actually is and where most people's understanding is is the biggest gap in business, ending with a train-and-station metaphor.

For brands planning a campaign

Useful when a partner needs to sound credible on timing to an enterprise or retail audience: the post shows fluency in the current protocol landscape.

• It also shows the sales register that comes with it: large, unlinked numbers and metaphor-heavy framing.
• Check how the underlying claims hold up before repeating them.

This is the set's most specific statement of the problem, so read it second when evaluating a partner.

Argues that the usual blockers—payments, identity and trust—are already solved: x402 ships stablecoin micropayments to any HTTP endpoint; Stripe and PayPal have agent SDKs in production; Google's AP2 launched with more than seventy partners signed for cryptographic purchase mandates; MCP gained OAuth 2.1 and dynamic client registration; and DIDs and verifiable credentials cleared W3C and now live inside eIDAS 2.0 wallets in the EU.

• It then names three real bottlenecks from inside the build.
• Cross-ecosystem trust: an agent running on Anthropic cannot act for a person whose other agent runs on OpenAI or a fine-tuned Llama in a private VPC, and no neutral registry vouches that an agent represents a named person with set spending limits.
• Dispute resolution: nobody has written the fiduciary-agent legal framework that decides whether a bad booking is charged back to the merchant, the model provider or the agent runtime.
• Adoption gap: MCP server adoption is growing fast, but almost none of it is customer-facing.
• Until consumer APIs such as Airbnb, Uber, Shopify and Booking wire up MCP or A2A with payment-ready mandates, agent-buys-for-you demos are theatre.
• At Rezolve, the work is the adoption surface, not new protocols.

For brands planning a campaign

The strongest scoping document here.

• It tells a platform exactly what it needs to supply before a pilot means anything: customer-facing MCP or A2A endpoints with payment-ready mandates, plus a position on liability.
• Ask the candidate directly about dispute resolution; the post shows the topic is already on their radar.

Open it for market sizing to justify a pilot internally and to see how the writer handles numbers when the headline figure cuts against the story.

Opens with AI-driven traffic to US retail sites up 3,300% year over year during Amazon Prime Day 2025, then qualifies the figure in the next line: that surge was still under 9% of total retail traffic, leaving 91% of shoppers searching, scrolling and comparing as before.

• Uses the jam study, where 24 options produced 3% purchases against 30% for 6 options, to argue that decision fatigue is a design problem scaled to a $6 trillion industry.
• Then cites Adobe Digital Insights for the 2025 holiday season: shoppers arriving from AI referrals converted 31% better, bounced 33% less and spent 23% more time engaging, while AI shopping satisfaction among Boomers rose 61% in a single year.
• Frames Rezolve's platform as removing decisions rather than adding options.

For brands planning a campaign

A useful internal-selling artefact and a test of rigour: the headline number is immediately paired with its own counterweight, a good sign.

• The Adobe figures sit behind a link in the comments, so verify them before reuse.
• Best fit for retailer-facing pitches where decision fatigue is the argument.

The shortest and least technical post here, making it the best first read for a colleague who does not follow protocol news.

Describes opening Amazon to buy protein powder, seeing 1,847 results, closing the tab and buying nothing.

• Contrasts roughly 7,000 products in an average 1995 supermarket with 350 million on Amazon, notes e-commerce conversion has stayed at 2-3%, and returns to the jam study and Barry Schwartz's 2004 book on choice.
• Argues the industry optimised checkout speed, UX, retargeting and button colours without questioning the assumption that more choice is better.
• The conclusion: the next era is no search at all—one sentence from the buyer, one right answer out, not 1,847.
• Closes by asking readers what they last did not buy because choosing felt like too much work.

For brands planning a campaign

The best piece to put in front of a merchant or category buyer who does not follow protocol news: it makes the case in their language and ends with an engagement question. It has no product detail or metrics, so pair it with a technical piece for an engineering or product audience.

Read it for the analogy, the clearest illustration of the concept in this set.

Recounts firing a Dragunov, a Desert Eagle and a Tavor at Battlefield Vegas.

• The instructor did not empty every weapon onto the table and tell the author to pick one; they watched the author's grip and stance on the first gun, then chose the next based on what they saw, personalising in real time with no menu and no filters.
• Sets that against online shopping that offers 4,000 products, sort by price, filter by colour, enjoy page seven, with nobody watching or adapting to what the buyer actually needs.
• Describes the Rezolve build as extracting maximum context from one sentence and serving the right product: a conversation, not a catalogue dump.

For brands planning a campaign

A ready-made demo or event script: the instructor-versus-catalogue contrast explains agentic shopping to a mixed audience in under a minute.

• It is positioning, not evidence, with no metrics and a vivid personal frame that will not suit every brand.
• Use it as the hook and follow up with the technical piece.

One opens with an abandoned Amazon search for protein powder — 1,847 results, tab closed, nothing bought — and sets it against the roughly 7,000 products in a 1995 supermarket, Amazon's 350 million listings, and e-commerce conversion rates still sitting at 2–3%.

• The other revisits the jam study (24 options on a shelf, 3% bought; 6 options, 30% bought) alongside a Prime Day 2025 figure of 3,300% growth in AI-driven retail traffic that still amounted to under 9% of total traffic.
• Both land on the same conclusion: the next step is not better search but fewer decisions.

A personal shopping failure hooks readers; option counts and conversion data turn it into an industry claim.

For brands planning a campaign

A partner can expect to be judged on whether it offers one answer or a better list. The most useful demo is a single resolved recommendation; a refined results page reads like the same old shelf with new filters.

Examples: Source 1 · Source 2

It closes by naming Rezolve AI as the place where this work is being built.

A second post argues the other half of the story: merchant-side wiring is largely missing. Readiness claims are therefore an active argument in the material, not a settled view.

For brands planning a campaign

The framing is post-infrastructure. An integration partner is measured on what it adds on top of existing protocols, not on proposing another standard.

Examples: Source 1 · Source 2

It lists working pieces — x402 stablecoin micropayments, Stripe and PayPal agent SDKs in production, Google's AP2 with seventy-plus partners, MCP with OAuth 2.1 and dynamic client registration, DIDs and verifiable credentials now inside EU eIDAS 2.0 wallets — then names three gaps: no neutral registry that vouches for an agent's identity and spending limits across ecosystems; no legal framework for charging back a bad agent purchase; and adoption concentrated in internal tool orchestration rather than customer-facing commerce APIs. The work is described as the adoption surface, not the protocols.

For brands planning a campaign

A vendor with something concrete to say about agent identity, dispute resolution, or consumer-facing merchant integration has an obvious entry point. General enthusiasm about agent commerce risks sounding like a stage Felix T. has already moved past.

Examples: Source 1

Standards and launches are named specifically — Verifiable Intent, Universal Commerce Protocol, Agentic Commerce Protocol, AP2, MCP, eIDAS 2.0 — rather than referred to generically. The underlying report is pointed to rather than reproduced, so the figures rest on the cited source.

Data-drop structure: a striking figure in the first line, a counter-figure immediately after, then the interpretation.

For brands planning a campaign

Partner claims land best with a citable figure or a named standard. Broad market-size assertions without a source are exactly the kind of claim these posts argue against.

Examples: Source 1 · Source 2 · Source 3

The clearest is a Las Vegas shooting range, where the instructor watches a first weapon, adjusts stance and grip, then picks the next one based on what they saw — set against a catalogue dump with sort, filter and page seven.

• The other uses the abandoned protein-powder search.
• In both cases, an abstract point about personalisation or choice is made physical first, then carried into commerce and, in the shooting-range post, into Rezolve AI's conversational commerce.

Physical anecdote first, commercial generalisation second.

For brands planning a campaign

A brand can supply a concrete scene or demonstration and expect it to become the argument. This style suits products that can be shown in one real-world moment rather than listed as features.

Examples: Source 1 · Source 2

Hashtags appear on some posts but not others. Rhythm and juxtaposition carry the argument, with little extended explanation in a single post.

For brands planning a campaign

Long-form co-created content would need editing to fit. The format rewards one strong opening line and one clear conclusion per post, making it a better fit for a single integration point or counter-intuitive finding than a full partner overview.

Examples: Source 1 · Source 2 · Source 3

Ask for their title, ownership stake, and employment terms.

Original disclosure

"At Rezolve AI we build commerce that works like that instructor." / "That's the wedge we work on at Rezolve." / "This is what we're building toward every day at Rezolve AI."

View example

Gai’s posts are retailer-facing explainers that list protocols and show how brands can become agent-ready. Felix Tseitlin’s posts are first-person arguments, using anecdotes and market figures to make the case about choice overload and adoption.

Read original · Original 2

Shivers explains how competing standards work and how retailers can prepare, including through structured feeds and checkout endpoints. Felix Tseitlin focuses on adoption and trust bottlenecks, and on product experiences that turn a need into one answer.

Read original · Original 2

Gerada’s posts diagnose processing rates and pitch Krepling Pay. Felix Tseitlin writes from a commerce-platform perspective at Rezolve AI, using consumer-shopping examples and arguments about protocol adoption rather than fee breakdowns.

Read original · Original 2

The two posts offer implementation guidance: one walks through OpenAI and Stripe's Agentic Commerce Protocol—what it enables inside ChatGPT and a five-step retailer list covering structured product feeds, checkout endpoints, delegated payments, and clean pricing and inventory logic.

• The other turns Google's Universal Commerce Protocol into a short action list: fix product feeds, strengthen the brand entity and schema, publish citable content, and make payment, fulfilment, and policies machine-readable.
• That is the right shape when the platform sells the integration, the audience is e-commerce operators who have to act this quarter, and the standard being adopted is already the platform's own.
• Felix T. argues the whole market's adoption gap and leaves the mechanics to the reader; Shivers starts with the mechanics.

The tradeoff is argumentative force.

• The reviewed work explains what a protocol does and what to fix first, accepting the standards as its frame rather than contesting them.
• A campaign handed to this writer still needs the platform to supply its position, the comparison with rival approaches, and the reason a merchant should choose this partner over another.
• The work also offers less of the sceptic-facing material Felix T. brings: no discussion of cross-ecosystem agent identity, chargeback liability, or who owns the customer relationship beyond stating that the merchant keeps it.
• Where Felix T. opens with a quantified hook and a signature line such as “the feed being the new homepage,” Shivers closes with an offer to share a checklist—useful for lead generation, but a thinner base for a point-of-view deliverable.

Examples: Source 1 · Source 2

The reviewed work breaks down a merchant's processing rate with worked arithmetic: an operator processing $110M a year believed they paid 2.9% + 30¢, but their blended rate came to just over 4.3%, or about $1.54M a year against gross margin.

• It then makes the agentic argument directly: when software initiates the transaction, software also reads the fee schedule, so opaque pricing stops being a moat and becomes a liability.
• The second post extends that into a position on consent, arguing that approval has to move from the individual transaction up to the relationship, and that merchants and agent developers will ultimately want a settlement layer with no stake in what gets bought.
• That is a ready-made case if the platform sells a checkout or payment layer and wants an argument about fee schedules and neutrality.

The same work puts a competing product in the frame.

• The posts describe Krepling Pay at a stated rate and a platform for the open web that picks no winner, so a campaign would have to explain why the platform's pricing or neutrality is different rather than simply borrow the argument.
• The diagnostic, arithmetic-heavy register will land differently with a retail-technology audience expecting the discovery and merchandising side of agentic commerce: neither post covers product feeds, search visibility, or how a shopper's request becomes one recommendation.
• Felix T.'s strongest ground is that readable consumer-to-practitioner bridge, so a campaign that needs both the cost case and the shopping experience cannot be carried by Gerada alone.

Examples: Source 1 · Source 2

One post surveys emerging agentic payment protocols in a numbered list—Google's Agent Payments Protocol, Mastercard's Agent Pay, Visa's Trusted Agent Protocol, PayPal's agentic commerce platform—and closes with the practical imperative for agent-ready brands to design storefronts, APIs, and product data so agents can understand, trust, and transact with them.

• The other moves from a consumer-behaviour example to how the different layers of agentic commerce will work, then points to a previous article on getting discovered by answer engines.
• That makes it the useful choice if the platform wants a single partner to hold the discovery-to-checkout story.
• It is also better for entry-level trend framing than for a technical audience.

The published pieces frame and survey rather than argue.

• The payment post lists four protocols without ranking or testing them, and asks what could go wrong in a rhetorical aside rather than answering it.
• The agent-ready-brands conclusion is a general call to action, not a recommendation backed by a decision.
• A platform campaign would therefore need to supply the sharper thesis, specific proof, and reason to choose one integration path—the work Felix T. does when they name cross-ecosystem trust, dispute resolution, and customer-facing adoption as the actual bottlenecks.
• The posts also shift between a consumer beauty anecdote and a B2B protocol rundown, so a practitioner audience may find the opening softer than Felix T.'s quantified example or Gerada's cost arithmetic.

Examples: Source 1 · Source 2