Everything in one place - writing, publications, projects, and the monthly roundup. Filter by tag.
AI providers ask us to trust agents with our accounts and real work, while their terms of service put every consequence on the user. Before I hand an agent the keys, one question needs an answer: who is responsible when something goes wrong?
Articles and tutorials on serverless technologies from Upstash and community
Durable execution engines compared in 2026: Upstash Workflow, Temporal, Inngest, Trigger.dev, AWS Step Functions, and Cloudflare Workflows, across architecture, retries, long delays, human-in-the-loop events, serverless support, and pricing.
Every major way to run background jobs, message queues, and scheduled tasks from a serverless app in 2026: QStash, Inngest, Trigger.dev, Google Cloud Tasks, AWS SQS and SNS, Cloudflare Queues, and Vercel Cron, compared on HTTP delivery, retries, dead-letter queues, delays, cron, and price.
We rewrite a model's own answer in the transcript to say the opposite, then ask it to explain itself. Across ten frontier models, 46% of replies defend the planted answer. Opus 4.8 almost never falls for it. Fable 5 does worse than both Opus and Sonnet.
Practical ways to cut Claude token costs in Claude Code and on the API: prompt caching, model tiering with subagents, effort tuning, context engineering, skills, and Context7. The techniques work on every model, from Haiku 4.5 to Fable 5.
Factory robots (1990-2015) are the best natural experiment we have for machines replacing workers. The mechanics of displacement carry over to AI, but the target inverts: robots hit the least-educated hardest, while AI exposure rises with skill and income. A data-driven look at who actually loses their job.
A monthly public record of the links I save - some of what I read this month.
The best managed Redis hosting providers compared on product, in 2026: Upstash, AWS ElastiCache and MemoryDB, Redis Cloud, Google Memorystore, Azure, DigitalOcean, Aiven, and Railway, across connectivity, global replication, high availability, and built-in search.
An end-to-end data platform you run on your laptop, built phase by phase on k3d: streaming ingest, row/column/lakehouse storage, observability, and resilience.
A local, end-to-end data platform on one k3d Kubernetes cluster, built on a laptop in tagged phases: streaming ingest, the right storage shape per access pattern, observability, and resilience.
Sakana's Fugu orchestrates frontier LLMs into a single endpoint. A guided tour of its two engines (the evolved TRINITY router and the RL-trained Conductor) with interactive diagrams of both architectures.
AI agent primitives on Upstash Redis: memory, RAG, search, chat history, caching, and rate limiting. With Vercel AI SDK and Eve
Notes from building a quick demo agent on Vercel Eve: initializing a project, the file layout, the dev TUI, tools and skills, scheduling, channels, deployment, and how it builds on Vercel's stack.
Ask HackerNews anything: a Vercel Eve & Upstash Redis Search demo agent.
How we gave a Vercel Eve agent retrieval-augmented answers by backing it with Upstash Redis Search: pass-through search tools and a system-prompt query reference.
Capture Vercel AI SDK generations and tool calls as JSON in Upstash Redis, then serve latency percentiles, token stats, and error counts entirely with Upstash Redis Search aggregations — no sorted sets, no client-side math.
Run durable AI agents on Upstash Workflow with @upstash/workflow-agents, then keep their payloads and token usage in check by splitting work across subagents with an orchestrator-worker setup.
Two approaches to building analytics on Upstash Redis: pre-planned counters with bitmaps, and flexible event recording queried with Redis Search.
The 2026 Upstash Redis vs. Valkey comparison. The fork, the hosting options, architecture, pricing and when to pick which.
My personal website: writing, publications, and projects. Built with Astro, TypeScript, and Tailwind CSS.
A personal site collects everything you do in one place. And you can build one without knowing a thing about the framework it runs on, just focusing on the aesthetics, like Bob Ross.
AI agent traffic analytics for Next.js
A monthly public record of the links I save - some of what I read this month.
Collection of skills for Upstash
A satirical stock-market/news simulation. CEOs resign over fishing rights, countries declare war over cheese tariffs.
A simple flight simulator written in Rust
A lightweight analytics platform built on Upstash Redis, providing event tracking, session management, feature flags, and A/B testing with a type-safe SDK and a built-in dashboard
TypeScript SDK and CLI for Upstash Box — sandboxed AI coding agents
Dynamically convert OpenAPI specs into AI agent tools
An extension of @upstash/workflow for building reliable multi-agent AI systems
How to back a Model Context Protocol server with Redis for low-latency, cheap tool calls.
Search Vercel Changelog and Blogs with Upstash Search
Patterns for modeling, writing, and querying time-series data on Redis.
Add a Redis cache layer to Drizzle ORM queries for faster reads in serverless apps.
Durable, long-running AI agents built on Upstash Workflow.
Practical tactics to cut Vercel function spend without rewriting your app.
Durable, Reliable and Performant Serverless Functions
Rate-limit LangChain apps and agents using Upstash Ratelimit.
Fixing Poetry's missing-python error on macOS without installing python-is-python3.
A RAG chatbot template built with the Vercel AI SDK, LangChain, Upstash Vector and OpenAI.
A tutorial deploying an Elixir/Phoenix app that caches WeatherAPI responses in Upstash Redis on Fly.
Building a retrieval-augmented chatbot over university data with Upstash Vector.
A visual particle simulation with Python and Turtle Library
AI chatbot for expert answers on university degrees
A fine-grained intrinsic and extrinsic comparison of contextual and static word embeddings for Turkish.
My solutions to the sample quantitative-researcher questions Optiver shares in their job listing.
Replicating win-stay/lose-shift strategy evolution in the iterated Prisoner's Dilemma, with Java multithreading.
Implementing the TabTransformer model in Keras, with an RNN warm-up, during a data-science internship.
Repository for assignments of the cybersecurity minor in University of Twente
A tool that generates goal models from agile user stories, letting teams benefit from both representations.
Chess engine and Minimax AI
Implementation of TabTransformer, attention network for tabular data, in Keras
Notebooks on time-series demand forecasting
Studying data science and machine learning at Arute Solutions as an intern
A Simple AI using Min-Max Tree Search
Game of Life simulation using Python and Turtle
Itearted prisonner's dilemma simulation and results
Beginner level mine sweeper using python and turtle
No articles match that tag.