Key Takeaways
- Tether's reliance on a single reliable banking partner creates a structural bottleneck, with limited exposure to a troubled offshore bank.
- Starlink adoption is rising in Europe, with users citing low latency and reliable service during ISP outages.
- Indie hacking is splitting into two viable paths: fast AI tools and niche, unglamorous products, with the middle ground disappearing.
- A data labeling firm predicts most future revenue will come from Fortune 1000 companies, not AI labs, and that evals will become key proprietary IP.
- A blue-collar booking widget founder reached ~$55K/month with no coding experience, using a simple stack and a sell-first approach.
- Acquire.com lists multiple small SaaS projects for sale, while data shows seed-stage unicorn odds are under 1% and founder equity dilutes sharply.
- Traditional enterprises face a mix of optimism, frustration, and fear over AI transformation, with many stuck at the coding agent stage due to missing eval infrastructure.
1. AI Models and Infrastructure
- Tether is again facing funding constraints, which Patrick McKenzie attributes to a structural need for two banks—one to custody the pool of funds and one for high-frequency transactions—while Tether currently has only Cantor as a reliable bank-like entity. A cited report says some Tether funds are stuck with an offshore banking partner facing liquidation risk after US asset seizures; Tether says its assets at EQIBank are limited, under 0.034% of total assets. 1
- A large data labeling company predicts that in a few years most of its revenue will come from Fortune 1000 companies rather than AI labs. It argues every company will want its own intelligence, but that does not necessarily mean using open-source models, and that as internal eval environments are built and agents improve, the company's evals will become its main proprietary IP. Most enterprises have not moved beyond coding agents, mainly due to a lack of suitable eval infrastructure to make non-engineering agents perform usefully. 1
- Ben Tossell argues dynamic model routing can create a more effective free market among model providers by routing tasks to competing models, pressuring providers on pricing and pushing them to offer the best, fastest, or cheapest tokens for a given task. 1
2. Connectivity and Independent Hacking
- Starlink adoption is increasing in Europe, according to @levelsio: he and two friends now use it, one friend in Portugal borrowed a Starlink Mini after ISP NOS was down for over two months—despite living 500 meters from NOS's Lisbon headquarters—and then ordered his own, while another friend in Romania just got one. He reports latency around 10ms, comparable to or better than fiber, no disconnections since use began, and customer support resolvable via chat. 1
- @levelsio describes the future of indie hacking as a barbell: one end is building AI tools for AI practitioners, which are easy to clone and won by fastest execution and distribution, with revenue that can swing wildly; the other end is boring, hyper-niche products nobody wants to clone, such as a CRM for Chihuahua owners or a communications tool for bakeries, requiring years of slow grinding to accumulate market share. The middle ground has become a dead zone. 1
- Tony Dinh compares TailwindCSS being chosen by AI as a "standard" and speculates Lego could become the physical world's TailwindCSS. A cited post describes Opus 5.5 being asked to design a Microduck from real Lego bricks: designed to physical scale using 1,113 real Lego parts, validated with 3,204 connections, 0 collisions, buildable step by step, center of mass within the feet, generating a 141-page Lego-style instruction manual with 237 steps, pricing each part in the browser, and preparing an order on BrickLink. 1
3. Startups, Sales, and Enterprise AI
- A founder with $140K monthly revenue checks his phone first thing every morning because he must confirm nothing is broken and no anomalies have occurred. 1 Tom, who built a small booking widget for blue-collar trades like electricians and mechanics, now earns AUD 85K/month (~USD 55K). He had no coding experience, asked a developer friend to build an MVP in February last year, and hired a lead designer, development lead, and front-end developer before having paying customers. His startup capital came from a previous marketing agency that still has over 70 clients paying for website hosting. He got first customers through door-to-door outreach rather than Twitter or UGC: the team walked into music studios, repair shops, and electrician shops, first showing the widget on an iPhone where nobody understood it, then switching to an iPad showing it embedded in a real brand's website, and handing out brochures. His customers—tradespeople who give direct feedback and won't pay for something that doesn't pay for itself—led him to focus on this group. Revenue grew from AUD 597/month in August last year to about AUD 63K/month in roughly a year. His stack is simple: AWS (150), Slack (120), Google Workspace (130), Vapi (500), Twilio SMS (4000), and Claude. His advice is to sell first, build later: if you can get someone to commit to paying before the product exists, you've likely found something worth doing. 2
- Alex Lieberman says traditional enterprises face three emotions around AI transformation: optimism from existing moats like brand trust, proprietary data, and network effects—unfair advantages disruptors haven't accumulated; frustration from slow AI tool adoption, difficult pivots, and a lack of urgency to become AI-native; and fear that AI transformation may take 5-10 years but they may not be able to wait and could lose market position. 1
- Acquire.com lists several SaaS projects for sale: a no-code data import automation tool with TTM revenue of $39.7K asking $137K; a digital business card SaaS for resellers/white-label clients claiming category leadership with TTM revenue of $100K asking $199.8K; a supplement brand with a no-inventory custom-label dropshipping platform with TTM revenue of $24K asking $77.5K; and an AI video studio that mass-produces branded short videos based on viral formats with TTM revenue of $22.7K asking $32K. 1 2 3 4 A cited post says a study of over 1,000 seed-funded startups shows less than 1% become billion-dollar companies; Carta data shows median founder team ownership is about 56% after seed, about 36% after Series A, and only about 16% after Series C, split across the whole founding team. The post argues that instead of chasing a billion-dollar valuation, bootstrapping, owning 80%-100% equity, reaching profitability, and selling for $3M-$5M can also deliver substantial wealth and time freedom while avoiding most stress; it acknowledges building a unicorn is still an impressive achievement and founders need not treat it as the only measure of success. 5
- Codie Sanchez says her company has stuck to no pivots and no restarts, operating Contrarian Thinking for six years, and lists Bizscout at 174K users, Contrarian Thinking having taught over 15K members, and Contrarian VC Fund as a top decile fund. She proposes 12 profit levers for a business covering pricing, customers, problems solved, how they are solved, how much you pay yourself, sales method, hiring, customer acquisition, running without you, which problems to prioritize, what you truly want from the business, and life outside the business; two are the biggest drags and should be fixed first. She argues that being able to explain a business does not mean you can run it, citing a former Starbucks CEO who spent 19 years at McKinsey and only 3 years running a company, after which market cap fell from $114B to $72B in 17 months; her conclusion is that the best business school is actually running a business. 1 2 3
