Written by Pedro Teixeira, a Software Engineer
An engineer-friendly deep dive into how the human brain truly processes information, learns without backpropagation, and uses spiking neural networks, predictive coding, and dopamine-driven TD errors—challenging our notions of AI and matrix math.
How we evolved our Lambda-based real-time AI streaming stack for SaaS: integrating streaming, credit management, and subscription-based throttling. Patterns, challenges, and lessons from Helpmaton’s production deployment.
How to implement tiered API throttling in AWS API Gateway using subscriptions, without requiring clients to send or manage API keys. Learn how to map subscriptions to usage plans with a custom Lambda authorizer for seamless and secure rate limiting.
How we give AI agents long-term memory without blowing the budget: time-stratified vector memory (six grains, progressive summarization) plus a graph knowledge base for structured facts, with bounded storage and semantic search.
How building Helpmaton — a robust infrastructure layer for AI agents—helped transform one-off scripts into reliable, context-aware, and continuously evaluated AI workflows for projects like TimeClout.
In a world kept afloat by edits and dreams, an anchor’s family stumbles through the forgotten tunnels beneath the Spire—where the discarded, the rewritten, and the erased fight to remember what is real.
How I built a fast, fully client-side semantic search system for AI customer service agents, using vector embeddings, Web Workers, and real-time doc updates — all without dedicated search infrastructure.
Announcing the private beta launch of TimeClout, the open-source, AI-powered tool designed to simplify and optimize shift scheduling for teams.
Turning 50, closing a startup, and taking stock of a 28-year journey in software - from distributed systems to open source, from startups to public projects. Looking back to move forward, with renewed curiosity for health, education, and meaningful work.
Reflecting on the journey of building—and saying goodbye to—Decipad, this post explores what's next: open-sourcing our work, lessons learned, and a personal look toward new challenges in health tech and education.
Reflecting on the limitations of spreadsheets and the quest for a more intuitive, human-centered approach to working with numbers, this article explores why new tools like Decipad are needed—and what challenges remain beyond the grid.
From real-time collaboration with CRDTs to a high-performance Rust computation engine, we're pulling back the curtain on the 9 biggest technical hurdles we overcame in Decipad's evolution.
Discover the key lessons learned from building—and ultimately closing—Decipad over five years. CTO and co-founder Pedro Teixeira shares hard-earned insights on product development, pivots, team building, and navigating the challenges of creating a new category.
A retrospective look at the five-year technical journey of Decipad, exploring architectural decisions, key challenges, and product pivots that shaped its evolution from a modeling tool to a collaborative data notebook platform.
An engineer-friendly introduction to Reinforcement Learning in Computational Psychiatry: how Q-learning works, how the brain might implement similar algorithms, and what this means for understanding mental health disorders.
An accessible, engineer-friendly introduction to Predictive Coding: how the brain predicts the future, why prediction errors matter, and what this means for AI, robotics, and intelligent systems.
A practical introduction to the Hierarchical Gaussian Filter (HGF) in computational psychiatry—how it models human perception, belief updating, and learning under uncertainty, with intuitive explanations for software engineers.
A software engineer's deep dive into the core computational models that power modern psychiatry, exploring how code, math, and neuroscience come together to decode the brain's mysteries.
A software engineer's journey through the 2025 Computational Psychiatry Course in Zurich, exploring how computational models and systems thinking can be applied to understanding the human mind and psychiatric research.
Why beautiful designs alone can't overcome the realities of complex systems — and why collaboration between designers and engineers is essential for building things that actually work.
Why over-engineering early is a trap, and how to build software with just enough architecture to get real results. Embrace the 'minimum viable architecture' mindset and learn practical, low-friction alternatives to heavyweight solutions.
While tRPC offers great developer experience with end-to-end type safety, GraphQL's client-side query customization provides architectural flexibility that shouldn't be overlooked. A deep dive into why GraphQL's advantages remain relevant.
How we built a CRDT compaction system for Decipad, merging Y.js updates to dramatically reduce storage costs and improve real-time collaboration performance in high-frequency editing scenarios.
A comprehensive exploration of Decipad's unit conversion system, detailing how the language integrates dimensional analysis, enforces unit correctness at compile time, and enables precise, error-resistant calculations for scientific, engineering, and financial applications.
Discover how Decipad's date system models time as intervals with built-in granularity, enabling intuitive operations like 'contains' and natural date arithmetic that matches human reasoning.
Discover how Decipad's lazy column architecture enables seamless analysis of massive datasets by combining lazy evaluation, chunked access, and async computation for performance and scalability.
How we designed and implemented a reactive, type-safe, spreadsheet-inspired language with real-time updates, dependency tracking, and incremental computation.
A deep dive into why date handling in programming is more complex than it seems, and how to approach time-based problems correctly.
Learn the essential patterns and strategies for building robust, user-friendly transaction systems that handle edge cases gracefully.
Discover the critical importance of offline-first architecture for modern applications, and learn how to design systems that work seamlessly regardless of network connectivity.
How we built an AI agent that can see and understand visual content, opening new possibilities for intelligent automation and assistance.
A comprehensive overview of how database technology has evolved from simple file systems to complex distributed architectures.
A journey through the evolution of data architecture, from centralized databases to local-first systems that prioritize user control and offline capabilities.
A deep dive into building scalable, production-ready AI chat services with real-time streaming using AWS Lambda, Architect, and Vercel's AI SDK. Learn about architecture, local development, deployment strategies, and cost optimization for modern AI applications.
Exploring the tension between immutable infrastructure and the need for adaptable teams in modern software development, and why both are essential for sustainable software delivery.
Discover how to create a fast, lightweight local development environment that closely mirrors AWS serverless infrastructure—without Docker or complex orchestration. Learn best practices, architecture, and tips for productive serverless development on your machine.
Learn how to automatically deploy every pull request to its own isolated AWS environment, complete with custom domains, SSL certificates, and full-stack infrastructure. This guide covers the architecture, CI/CD pipeline, and best practices for building robust, production-like PR deployments that supercharge your team's development workflow.
A practical implementation of bringing data closer to edge compute, with real-world examples and performance benchmarks.
Exploring the challenges and possibilities of bringing data closer to edge compute, and imagining a future of non-uniform data access for distributed systems.