Thoughts

An in-depth exploration of blockchain fundamentals, focusing on consensus protocols, Byzantine fault tolerance, and state machine replication (SMR). Key topics include the blockchain stack, digital signature schemes, the Byzantine Broadcast Problem, and protocols like Dolev-Strong. Discusses vulnerabilities of Lazy SMR and Round Robin protocols and previews advanced topics like FLP Impossibility, Proof-of-Work, and Tendermint.

Exploring dating through the lens of game theory, this article examines the 37% Rule from the Secretary Problem. Simulations reveal the trade-offs between patience and risk in sequential decision-making. Practical takeaways: start early, avoid settling too soon, reassess often, embrace uncertainty, and focus on getting shit together to increase compatibility and opportunities.

Built a weekend text-to-speech app inspired by Speechify, focusing on efficient information absorption without feature bloat. Features include link-based audio generation at variable speeds using superior text-to-speech models. Planned upgrades include highlight syncing, voice commands, and near-instant audio streaming.

Sequoia’s PMF framework defines three archetypes: Hair on Fire (urgent needs, fast solutions), Hard Facts (shifting habits, redefining norms), and Future Vision (paradigm shifts requiring persistence). Examples include Stripe, Slack, and Nvidia. Achieving PMF requires iteration, execution, and timing—factors often outside rigid frameworks.

Crypto protocols ($2.39T market cap) and Web3’s composability aim to disrupt centralized platforms by empowering user ownership and innovation. Examples include on-chain scientific replication, grassroots product development, and blockchain-based autonomous worlds. Web3 fosters open ecosystems through dApps, enabling collaboration, seamless integration, and participant-driven evolution.

Direct Atomics built an equivariant GNN to accelerate DFT, reducing simulation time from months to minutes. Challenges included niche markets, resistance to drag-and-drop tools, and in-house/open-source competition. Mistakes—poor market validation, premature scaling, and overfixation on competitors—highlighted the need for user-driven design, minimal viable builds, and assumption testing to succeed in material science.