Andrés Morales, author at El Solitario

Open source maintainers reviewing code without getting paid for their work

Programming

60% of open source maintainers don’t get paid: the proposal to force payment

60% of open source maintainers don’t get paid for their work, and React, Elasticsearch, and Redis found that out when they tried to change licenses and lost to a free fork. An essay from Seldo.com proposes forcing payment through registries instead of licenses.

By Andrés Morales, 3 weeks20 September, 2026 ago

Diagram of the rsync algorithm with block-level checksums

Programming

Rsync: The Algorithm That Only Moves What Changed in a File

Rsync syncs files without retransmitting them in full, thanks to a block-signature algorithm and rolling checksums. Learn how it works under the hood, with code examples and when to use it over alternatives like scp or Borg.

By Andrés Morales, 4 weeks20 September, 2026 ago

Brood War Bench leaderboard showing AI models playing StarCraft

Tech News

Brood War Bench: Codex Astra Wins 100% of Its StarCraft Matches

Brood War Bench, a new benchmark, tested Codex Astra, Claude Fable, Grok 4.6, and other models playing full matches of StarCraft: Brood War. None beat beginner level, though Codex Astra won all 18 of its matches without a single loss.

By Andrés Morales, 4 weeks20 September, 2026 ago

AI-generated event poster in geometric Bauhaus style

Tech News

ChatGPT Offers 12 Design Styles for AI-Generated Posters

A developer proved that ChatGPT can break out of the generic AI-generated poster look just by changing the prompt. The result was a list of twelve reusable design aesthetics, from Bauhaus to risography.

By Andrés Morales, 4 weeks19 September, 2026 ago

Diagram of HTTP caching with Cache-Control and ETag between browser and server

Programming

Cache-Control and ETag: How the Browser Avoids Asking Again

Cache-Control and ETag are the two headers that decide whether the browser reuses a response or asks for it again. This guide explains each directive with real code, a diagram of the cache hierarchy, and the mistakes that most often break production.

By Andrés Morales, 4 weeks19 September, 2026 ago

Laya AI decision engine processing text in milliseconds

Programming

Laya: AI Decision Engine at 32.8ms, Up to 8 Times Faster Than Jev

ConvAI Innovations introduced Laya, an open-source AI decision engine that responds in 32.8 milliseconds per query. The launch comes weeks after TypeSafe AI, founded by a former OpenAI collaborator, released a commercial product with an almost identical idea. Unlike Jev, Laya publishes its weights, code, and datasets under the Apache 2.0 license.

By Andrés Morales, 4 weeks19 September, 2026 ago

Claude Code terminal screen showing AGENTS.md support

Artificial Intelligence

Claude Code 2.1.277 Adds Support for AGENTS.md as an Alternative to CLAUDE.md

Claude Code 2.1.277, released on September 18, 2026, adds support for AGENTS.md as an alternative to CLAUDE.md when the latter doesn’t exist in the project. The release also fixes more than twenty-five bugs and adds controls for corporate gateways.

By Andrés Morales, 4 weeks18 September, 2026 ago

Artificial Intelligence

Andrew Ng Calls AI Extinction Risk Science Fiction

Andrew Ng, co-founder of Google Brain, called warnings about human extinction from AI science fiction. The debate pits those pushing to pause the largest models against those who consider that risk technically implausible.

By Andrés Morales, 4 weeks18 September, 2026 ago

Database sharding diagram with partitions distributed across multiple servers

Programming

Sharding: How to Split a Database to Scale Horizontally

Database sharding splits a giant table into independent fragments that live on separate servers. This guide explains hash, range, and directory-based strategies, with examples in PostgreSQL and application code.

By Andrés Morales, 4 weeks18 September, 2026 ago

Source code illustrating vibe coding and formal software verification

Tech News

Bend 2 Needs 442 Lines to Prove What SPARK Solves in 12 Checks

Bend 2 presents itself as the ideal language for the AI era: humans write laws, AI writes the proofs. But to prove a player never touches a flag it needs 442 lines of proof, while SPARK, a formal verification language with decades of history, solves the same thing in 12 automated checks. The case exposes a classic vibe-coding trap: building a lot before checking whether a better solution already exists.

By Andrés Morales, 4 weeks18 September, 2026 ago

Andrés Morales

Developer and AI researcher. Writes about language models, frameworks, developer tooling, and open source releases. Covers ML papers, the tech startup ecosystem, and programming trends.