Stratechery — Ben Thompson’s technology-strategy writing
Stratechery — Ben Thompson’s technology-strategy writing
Subscription. Dru added Stratechery on October 2, 2026 with every item delivery to Agentic Software. Every new item receives a link post, even roundups and schedule notices. Always use a canonical source URL without the subscriber feed's bearer access token. Originals combine observed agent use, announcement and strategic inference; keep these epistemic statuses separate. Agent as interface and service, disposable prototype versus production, and human orientation are the main threads. Not independent comparative field evidence.
September 25–October 2 intake. Aggregator Wars (recap), Apps, Agents, and Aggregation (custom personal UI), Meta Enterprise (personal/work identity), Dot (cross-project chief-of-staff hypothesis), weekly recap. Five hearts, including four originals not opened as sources; do not treat hearts as full-article endorsement.
October 12–19. October 12 Personal Day is a schedule notice; October 13 SpaceX distributed-cellular thesis is off-programme; October 14 HomeView/App Story pits integrated accessories against the iPhone/voice boundary; October 16 Home At Last is a recap, not deployment; October 19 Do You Trust Apple? frames integration, support and Siri as a household commitment. No actual household use yet. Each was delivered owing to every-item preference.
October 20 original, subscriber feed received in full: Muse Spark Open, Meta + Apple, Thin and Thick moves from Meta’s restricted open-weight launch to Thompson’s witnessed two-Mac Apple coding/video demonstration, then proposes speculative branches and prebuilt follow-on feature paths if local model tokens are cheap. Research note checks model, license and MLX benchmarks against Meta and Apple, and compares Cursor’s already available independent best-of-N worktrees. The proposed branching harness was not the coding demo, and inference metrics do not establish all-in accepted-feature economics. The canonical page verifies the title/date but full subscriber text came from the item; do not copy that text into public notes.
October 21 original: Netflix earnings, business or franchise, Disney’s deal argues that retention and pricing strength lag the question of whether original content keeps growing engagement; the Disney license still showcases Netflix’s distribution. No agentic software evidence; relevant only as an explicit proxy-versus-actual-demand analogy. Full text received from subscriber feed; canonical link checked. Do not conflate advertising shortfall, which management also attributes to partner integration, with a measured viewing decline.
October 21 later original: Presented by Stripe introduces an exclusive sponsor on top of subscriptions, ads even for paying readers, an explicit ethics disclosure and a claimed editorial-independence clause. Thompson says revenue will support more products, beginning with video Sharp Tech and an additional subscriber audio show. This is a business-model disclosure, not an agentic-development observation; no measured subscriber benefit or independence test yet. Posted under Dru’s every-item preference with narrow product-and-attention relevance, no extra discretionary post.
October 23 video item: Sharp Tech on Video announces the first video version of the podcast, offered on YouTube and on Stratechery; a subscriber-only Sunday episode is promised. The subscribed item and the canonical public title/date were checked separately; the public page hides the full item behind a subscription. The title mentions Netflix, the internet and Meta’s VR glasses, but the announcement does not develop an argument about them. This fulfills part of October 21’s announced product expansion, not evidence of audience preference, subscriber benefit or coding-agent outcomes. Do not mistake an episode announcement for the episode itself. Netflix/product signal analogy.
October 23 weekly roundup — brief and check. Question: does 2026.43: A New Era add a new result on agent-built software, or only summarize earlier arguments and a product launch? Check sponsorship/video promise, Netflix’s claimed content-versus-distribution problem, and Apple/Meta AI headline against what this item actually states. The subscribed full text and canonical public page agree: Thompson introduces Stripe sponsorship and video Sharp Tech, Sharp revisits Netflix’s content execution and growth challenge, and the rest links Apple/Meta edge AI, a film review, podcasts, and the already posted originals. The roundup includes a repeated sponsor disclosure, not a new experiment. There are no viewing, retention, subscriber-value, workflow or requester outcome measures. Compare the October 21 Netflix original and feature-to-outcome question before reading this as more than a weekly index. The documentary recommendation is not agentic-software evidence.
October 26 original — smallness as software business thesis. The End of the Internet argues that public blogging moved to paid niches and public social interaction to trusted group chats; in his analogy AI security and economics now drive software toward private tools for individuals and small teams. He reports private Stratechery podcast/task/calendar tools and personal household/sports tools, but no use, code, maintenance or financial ledger. The prediction of many new local markets is a thesis. Audit checks his PHP trigger against the official release announcement and contrasts one separate company’s six-week first use versus seven-month parallel cutover. Stratechery’s original is public at the canonical URL; no token needed.