# The Falkster Corpus: Building a Company in Public The Falkster.AI build, written as it happens: architecture calls, pricing experiments, what failed, and what the numbers actually were. The claim: A founder writing while building has information nobody else has. The value is in publishing the decisions before the outcome is known, including the ones that turn out wrong. Source: https://falkster.com/corpus · Built 2026-09-19 · Author: Falk Gottlob Contents: 4 pieces on this topic ## How to use this Paste this file into your assistant's project knowledge (Claude Projects, a ChatGPT project, a Cursor rule file, an AGENTS.md), then work normally. The point is not to ask it about the corpus. The point is that when you ask it to size a bet, write a brief, or decide what to kill, it answers the way this practice answers instead of the way the average of the internet answers. Every entry carries a canonical link. When something here matters to a decision, follow the link and read the argument. A summary is enough to act on and not enough to disagree with. ## Attribution Written by Falk Gottlob. Free to use for your own work and your team's. When it shows up in something public, cite it as: Falk Gottlob, falkster.com, with the canonical link. Not licensed for republication, resale, or model training. --- ## Everything on this topic (4 pieces) Each entry is the piece's own extractable summary. Follow the link for the argument. ### Salesforce Named the Harness. The Missing Piece Isn't on the Slide. Published: 2026-09-14 Canonical: https://falkster.com/blog/salesforce-named-the-harness Salesforce's enterprise AI harness names six capabilities, Trusted Context, Agency, Action, Governance, Security, and Models, with an AI Control Plane above them, composable packaging, and consumption pricing with no number attached. Kumar volunteered on the press call that security and the FinOps piece still need building. The gap that matters isn't the February 2027 date, it's that context is the only layer in the harness captured at write time, by the person making the decision, at the moment they make it. Everything else can be bolted on later. Intent can't. That turns a 2027 procurement story into a decision you are already making this quarter, whether you know it or not: where your team's reasoning gets written down, and whether that record is structured enough for an agent to read. ### Comment With a Frame: The Only Move That Compounds Published: 2026-07-10 Canonical: https://falkster.com/blog/comment-with-a-frame Commenting with a frame is the only move in public discussion that compounds: concede the strongest objection in full, then add a frame the thread did not have before you arrived. Applause adds nothing and dunking wins the exchange while losing the room. Two threads show the method. I conceded to Artem Ermakov that prose and eval are source and compiled artifact, then reframed it as relocating the source next to the code as a docstring rather than deleting it. I conceded to Marton Gaspar that you cannot prototype fifty ideas, then reframed the surviving bet as designing the tournament instead of picking winners. The frame is the asset, conceding is a signal rather than a cost, and the thread is the test. Both of this week's posts on this site started as comments. ### Falkster's Messaging Gateway: A Prototype-First Ship Story Published: 2026-07-05 Canonical: https://falkster.com/blog/falkster-messaging-gateway-ship-story Falkster needed to talk to people where they already are, Signal, WhatsApp, and Telegram, without forcing them into a new inbox. The enterprise-default answer is three bespoke integrations, each with its own auth, rate limits, and failure modes, scoped in a PRD and handed to an integrations team. We went the other way: one Matrix homeserver as a single bridging layer, with bridges out to each platform, prototyped and tested against real accounts within days. Building it surfaced the real problem, which the brief could not have: bridge reliability varies wildly by platform, so the actual work was not integrate with three APIs, it was build a guard layer that behaves consistently even when the bridges do not. The spec you can write before you build is always thinner than the one you discover while building. ### What Competing Against Claude Taught Me About Falkster Published: 2026-07-03 Canonical: https://falkster.com/blog/what-competing-against-claude-taught-me I ran a structured comparison of Falkster against the Claude ecosystem, not to build a battlecard but to find the gaps, and four came back sharp enough to turn into tickets the same day. Memory maturity is the most uncomfortable: Anthropic's cross-session memory is more mature, and you close that by living with the hard tradeoffs long enough to get the defaults right, not by copying the surface. Workspace surface area is a real tradeoff: Claude spans many surfaces, Falkster is concentrated, so we compete on depth in fewer places. Long-running agents is the gap we can close fastest, because self-hosting gives us a control surface for trust infrastructure a hosted competitor cannot match. Distribution is the gap with no clever answer: Anthropic ships to a default-installed base and we do not. Publishing this is a choice, and I think being specific about where we are behind is worth more than staying vague.