saas

Bilingual

Things I would have done differently (consulting)
2 months ago
- Investing in tools earlier, even for small monthly fees, saves time and money in the long run.
- Manually handling invoices and time tracking was inefficient; switching to a $40/month SaaS tool reduced invoicing time from 2-3 hours to 30 minutes.
- A $1k/year time tracking tool immediately highlighted missed billing that paid for three years of cost in the first month.
- Building reusable internal tools has provided ongoing benefits both financially and for team morale.
- Trusting gut instincts about clients, projects, or teams is usually correct about 80% of the time and should be heeded.
- Avoid building custom versions of existing SaaS tools unless the cost justifies it, such as for very expensive subscriptions over $200/seat/month.
Figma's woes compound with Claude Design
2 months ago
- Figma's initial product revolutionized browser-based design using WebGL and asm.js, overtaking Sketch as the market leader for UI/UX design.
- Figma's collaborative features drove adoption beyond designers to developers, PMs, and executives, but this non-designer user base now makes the company vulnerable to LLM-based tools.
- AI, especially LLMs, has become surprisingly competent at basic design tasks, reducing the need for non-designers to use Figma for things like reports and presentations.
- Figma's AI product, 'Figma Make,' remains underwhelming, while Anthropic's 'Claude Design' directly competes by enabling rapid prototyping from existing design systems.
- Figma's reliance on Anthropic for AI inference creates a conflict of interest: Figma funds a competitor while using a weaker model (Sonnet 4.5) versus Claude Design's Opus 4.7.
- Building products like Claude Design requires far fewer people than traditional SaaS companies, highlighting structural inefficiencies in firms like Figma with ~2,000 employees.
- Traditional SaaS moats (multiplayer, plugins, SSO) are less effective against AI agents, making companies like Figma more vulnerable to disruption by inference providers.
SaaS is not dead
2 months ago
- B2B SaaS is not dying; claims of its demise are exaggerated by people without real business experience.
- Businesses buy SaaS to offload non-core competencies, valuing focus on their own expertise over building and maintaining tools.
- The classic 'build vs buy' decision remains unchanged: building in-house has upfront costs plus ongoing maintenance, often exceeding SaaS costs.
- Critics often undervalue their time, spending weeks to build replacements for small monthly savings, which is impractical for real businesses.
- Doom-predicting about SaaS is often engagement farming or fear-mongering for profit, similar to past crypto hype.
- Real businesses are increasingly choosing to buy SaaS to let teams focus on core tasks, as seen with Keygen's 40% revenue growth and low churn rate.
Wall Street just lost $285 billion because of 13 markdown files
2 months ago
- Wall Street lost $285 billion after Anthropic launched a legal tool consisting of markdown files, highlighting the fragility of SaaS valuations.
- Agentic workflows can bypass traditional SaaS by working at a higher level, using source material to answer complex questions rather than just automating processes.
- Systems of record with robust API access retain moats in an agentic future, but legacy platforms with poor APIs face customer migration.
- The future of software may be headless, API-first solutions built for programmatic access, enabling agents to use data flexibly without traditional UIs.
SaaS Still Has a Moat. Just Not the One We Think
3 months ago
- SaaS's primary competitive advantage is not data lock-in but rather the learned behavior and habitual use of its user interface, which creates significant switching costs for companies.
- The data moat is weakening due to AI-driven data portability, agent-to-agent communication reducing reliance on human UX, and new data paradigms like context graphs and vector stores.
- Five headwinds threatening SaaS dominance: AI forcing data portability, agent-to-agent interactions diminishing UI relevance, new data paradigms making old structures a liability, internal context layers reducing individual SaaS tool necessity, and AI coding flooding the market with cheaper alternatives.
- Large SaaS companies should respond by building robust APIs and MCPs, owning the context orchestration layer, investing in AI-native data backends, and making it easy and affordable for companies to run processes on their platforms.
- Companies tired of SaaS sprawl should start by building a context layer that syncs with existing tools, create agentic workflows on top of it, and gradually phase out SaaS tools as the context layer replaces their value.
- Startups competing with big SaaS should focus on owning the new context layer, syncing with incumbents' data, and building innovative interfaces and capabilities that incumbents cannot easily replicate due to their legacy constraints.
- SaaS is not dying imminently, but pressures are compounding; companies that prepare for an agent-to-agent norm, context layers as systems of record, and AI-native data structures will gain a long-term advantage.

Things I would have done differently (consulting)

Figma's woes compound with Claude Design

SaaS is not dead

Wall Street just lost $285 billion because of 13 markdown files

SaaS Still Has a Moat. Just Not the One We Think