8 min read · 12 sources

Apple just redrew the commission chart for every iOS developer selling into the EU, and the new numbers only matter because a single app can now mix Apple IAP with alternative payments for the first time. Effective October 1, the Core Technology Fee dies in favour of a 5% Core Technology Commission on out-of-App Store digital transactions. Standard in-app purchase commission lands at 26% (15% for small business, mini, and video partners), alternative payment processing at 20% (10% for eligible programs), and a link-out path at 15% (10% for programs). Initial acquisition and store services fees are gone. Once you pick a lane, you stay in it for 12 months.

An accelerator writing checks at $330K to $2.5M post only needs a $33M to $250M exit to work; your seed at $5-10M post requires belief in a $500M to $1B outcome.

Apple rewrites EU App Store terms

The new structure is a single set of EU business terms developers can sign now, replacing the patchwork that the DMA forced through 2024 and 2025. Apps distributed outside the App Store still need notarization. The 5% Core Technology Commission is the number to watch: it taxes revenue Apple never processed, which is a different kind of leverage than the per-download CTF it replaces. For a developer processing €1M in external payments, that is €50K of new margin or new cost depending on which side of the table you are on. The 26/20/15/5 stack also creates a clear pricing menu for routing decisions, and the 12-month lock-in makes the choice sticky.

a16z on the Borderless Founder

Source: a16z.news ↗

Andreessen Horowitz published a long piece arguing the founders most likely to build the next category leaders keep one foot in their home country and one in the Valley. The mechanism is compounding: the founder knows the strongest engineers locally before the rest of the market catches on, lands enterprise customers at home for early proof points, then borrows Valley credibility to scale. Marc Andreessen’s old compounding-resources framing is the underlying pitch. The subtext is that a16z wants to be the connective tissue for these founders, offering visa sponsorship, intros, and crash space in exchange for deal flow the rest of the market will not see for another 18 months.

Linear Agent gets environments and browser use

Source: linear.app ↗

Linear shipped the upgrade its agent was missing: the ability to set up, run, and test code rather than just suggest diffs. Coding sessions detect the repo’s toolchain, install what is missing, and run the app. Linear Agent then drives a browser to verify the work where users actually experience it, capturing before-and-after screenshots and re-running tests if a check fails. Sandboxes are billable in 20-minute blocks at $0.25, and model usage is passed through at provider rates with no markup. Admins can cap spend per user or per workspace, reset daily, weekly, or monthly. For teams that have been reluctant to let an agent touch a real environment, the new pricing plus the browser-verified feedback loop is the missing piece.

Slack Code turns channels into task rooms

Source: slack.com ↗

Slack launched Code Channels: temporary rooms scoped to a single task where agents start working the moment they are mentioned, pull in team context, and archive when the work ships. The channel closes; the context does not, and stays searchable alongside the rest of your Slack history. Agents live in a new Agents tab, run quietly in the background, and inherit Slack’s enterprise security stack: EKM, DLP, Discovery APIs. The product is built on the GitHub Copilot integration with Slack as the messaging layer, which means the “I don’t code” employee can describe what they want, watch the agent work, leave inline comments, and approve a preview without ever opening an IDE. It is the same pattern Linear and GitHub have been pushing, just with Slack as the front end.

Lovable's brand, told by its first designer

Source: review.firstround.com ↗

The Firsthand account from First Round is the most concrete growth story in the newsletter. Lovable launched in November 2024, the designer (employee four) prototyped the original visual identity in 30 minutes, and the product hit $10M ARR in 60 days. The pink-gradient heart logo that became the brand’s signature was a second pass, rebuilt after user research showed visceral emotional reactions to the product. The point of the piece is not the logo but the operating tempo: small team, complete trust, fast iteration. For founders wondering whether brand still matters when the product is an AI app builder, the answer the Lovable team gives is that the brand is the thing people screenshot.

Commenting as a sales motion

Source: startupgtm.substack.com ↗

Startup GTM makes the case for commenting on prospects’ posts as outbound that does not ask for anything. A well-placed comment is a direct touchpoint with the prospect and free distribution to everyone in their network, which is the exact persona you are trying to reach. The author flags the most common mistake: targeting prolific posters. Posts-per-week is a vanity signal. Buying pressure is the real signal: contracts expiring, recent funding rounds, leadership changes. The recommended filter is “would I call this person right now,” not “does this person post every day.” One good comment on a contract-renewal announcement outperforms twenty comments on industry thought-leadership threads.

Why fundraising gets harder every round

Source: thefounderplaybook.hustlefund.vc ↗

Hustle Fund’s Founder Playbook walks through the arithmetic that makes Series A feel impossible even when the seed went smoothly. Early-stage investors need 100x because most of the portfolio dies before product-market fit, so one winner has to cover the rest. An accelerator writing $330K to $2.5M post-money checks only needs a $33M to $250M exit. A seed at $5-10M post underwrites a $500M to $1B outcome. A Series A investor underwrites a path to roughly $100M ARR, which is why $1M ARR is no longer enough. Entry valuation matters because a 10x outcome on a zero-return portfolio does not break even; only a 100x changes the math. The escape valves named are customer-funded growth, capital structures that do not require 100x (revenue-based financing, non-dilutive debt), and staying private longer.

Code is not revenue

Source: mironov.com ↗

Max Mironov’s “Technical Ingredients, Revenue Ingredients” is the sharpest read in the newsletter for anyone shipping AI-built software. He separates three things: executable code (runs without crashing), commercially deployable software (meets a paying customer’s expectations for uptime, security, and support), and a revenue-ready product (everything needed to generate ongoing revenue). Commercially deployable means SRE infrastructure, monitoring, failover, SLAs, data security, access management, automated test suites, help systems, and support. For AI products it also means model decay and data drift tracking. Mironov’s argument is that the “how fast can we ship with AI” conversation has been about executable code, and prototypes are not products. CEOs building with agents need to identify the next revenue to bottle before code development outruns selling.

Why AI diligence is different

Source: x.com ↗

A thread from Mardehaym (who embeds AI-native engineers into PE firms) argues traditional SaaS diligence fails on AI companies. The proprietary technology is often an API call to someone else’s model behind a frontend, and a pitch deck is a configuration file. The case study cited: a company claiming a “proprietary model architecture” whose codebase contained gpt-4o calls. The recommendation is to open the codebase and score model dependency, data moat, talent concentration, and substitution risk before assigning a premium multiple. ARR and growth do not reveal whether the moat is real. For acquirers paying AI multiples, code-level diligence is no longer optional.

Sales reps are specialists, not generalists

Source: saastr.com ↗

Jason Lemkin pushes back on the “great rep sells anything” myth. Reps specialize by motion (inbound vs outbound), brand context (category leader vs challenger), support structure, competitive environment, product complexity, and price point. A rep who closes $3-10K deals in three calls is not going to close a $250K enterprise deal that takes nine months and a security review. The rule of thumb: hire reps who have sold something harder than your product. A rep who closed $1M ACV deals is calibrated; a rep who has only sold $50K is not ready to run your $500K motion even if their last quarter was their best.

Meta's Startup School

Source: wersm.com ↗

Meta launched a three-month Startup School for 200 early-stage consumer brands, with VC and industry mentorship. The framing is growth partner rather than ads platform, and the bet is that habits formed now (audience definition, creative discipline, measurement) become paid-media behaviour later. The positioning is both defensive (against TikTok, Shopify, creator newsletters, and retail media networks all chasing the same brand budgets) and offensive (leveraging Facebook, Instagram, WhatsApp, and Threads scale to be the default growth stack before the brand picks a competitor). For consumer founders, the offer is free support; the implicit cost is alignment with Meta’s measurement and creative conventions.

AI adoption as an operating model

Source: edisonpartners.com ↗

Edison Partners argues enterprise AI has moved past “who is using AI” to “how are we using AI across the organisation.” The warning is concentrated adoption: engineering and marketing race ahead while finance and customer support sit untouched, which creates alignment gaps and teams running at different speeds. The recommendation is to require every function to tie AI initiatives to a real business outcome rather than treating it as a showcase. The downloadable 2026 Growth Index is the framework. For technical leaders this is the org-design version of the platform-engineering conversation: the question is no longer whether to ship AI, it is how to make every team competent at it without burning the budget on a hundred pilots.