iiinova — where your digital self keeps evolving

Antón @anton.galiglobal.com · 15/05/2026

Building Coral has kept me busy lately. Coral is a local-first SQL runtime for agents that lets them query APIs, databases, files, and other data sources through a single interface. Today I published my first article about how I use Coral to work on Coral: withcoral.com/blog/debuggi...

withcoral.com

Debugging Coral With Coral — Coral Blog

Using a Coral-powered code agent to query Coral's own OpenTelemetry data, connect it back to the Rust source, and close the loop on real debugging work.

130

Antón @anton.galiglobal.com · 26/03/2026

I'm using openobserve locally to monitor code agents and some internal tools. It works pretty well, it's a single binary and it has plenty of features (but it's not bloated yet)

040

Reposted by Antón

Almog @almog.xyz · 23/03/2026

Databases have strange economic incentives that end up making them "worse" over time. The problem is an asymmetry between who benefits from innovations at first (database companies) and who benefits in the long run (hyperscalers that sell hardware). Blogged: www.bitsxpages.com/p/the-broken...

bitsxpages.com

the broken economics of databases

Why database companies charge so much, earn so little, and keep making things complicated

181

Antón @anton.galiglobal.com · 18/03/2026

It's so cool to see a company in my area collaborating with the ATmosphere!

091

Antón @anton.galiglobal.com · 13/03/2026

More Rust stuff, today is about test harness crates. Plenty of options 🤔 docs.rs/trycmd/lates...

docs.rs

trycmd - Rust

Snapshot testing for a herd of CLI tests

000

Antón @anton.galiglobal.com · 12/03/2026

Pro tips for debugging compilation times in GitHub workflows: - Use `cargo build --timings` to generate an HTML report. - Use the new GitHub feature to attach it as an artifact and browse it directly in GitHub. - Update Swatinem/rust-cache and check the logs github.blog/changelog/20...

040

Antón @anton.galiglobal.com · 09/03/2026

The goal of this tracker is to detect statistically significant degradations in Claude Code with Opus 4.6 performance on SWE tasks. TL;DR Agents have bad days too marginlab.ai/trackers/cla...

marginlab.ai

Claude Code Opus 4.6 Performance Tracker | Marginlab

Track Claude Code's daily performance on SWE-Bench-Pro. Monitor for degradation with statistical significance testing.

130

Antón @anton.galiglobal.com · 07/03/2026

I attended with my daughter the Oracle4Girls event in the local computer science university today. We need more events like this. Huge thanks to all the volunteers who made it possible ❤️

020

Antón @anton.galiglobal.com · 06/03/2026

Took a Python DNA sequencing utility and rewrote it in Rust. Outcome: ~50× faster than CPython and ~20× faster than PyPy… in one shot antonmry.github.io/FastFiRSt/be...

150

Antón @anton.galiglobal.com · 05/03/2026

Code reviews are the new bottleneck and not a pleasant one. Pushing reviews toward LLMs, specs, and BDD seems sensible, but I still feel friction. It could be immature tooling, but I suspect there’s something deeper at play www.latent.space/p/reviews-dead

latent.space

How to Kill the Code Review

Human-written code died in 2025. Code reviews will die in 2026.

120

Antón @anton.galiglobal.com · 04/03/2026

It was harder than expected but I've just migrated to @eurosky.social 🎉

021

Antón @anton.galiglobal.com · 02/03/2026

Connecting my code agent harness to Discord turned out to be surprisingly effective. I can move between phone and terminal, track long-running tasks via notifications, and isolate projects in separate Discord channels github.com/antonmry/bre...

github.com

Release v0.2.1 · antonmry/breo

Features Streaming Discord responses — Long-running LLM tasks now show live progress in Discord: an ⏳ Working... message appears immediately and updates every 3 seconds with the latest output, the...

000

Antón @anton.galiglobal.com · 28/02/2026

Engineers using LLMs to summarize design documents written by other engineers with LLMs. This is not the way. Today, more than ever, less is more.

040

Antón @anton.galiglobal.com · 26/02/2026

It's surprising how much Claude Opus 4.6 struggles with git rebases and merge conflicts

120

Antón @anton.galiglobal.com · 25/02/2026

I’m moving away from cargo-release with git-cliff. Good tools, but the setup friction doesn’t quite fit my workflow. Now testing: manual GH Action for releases and a bash script with `claude -p` to draft notes, quick pass in $EDITOR, publish with `gh`. Much simpler so far and better results.

010

Antón @anton.galiglobal.com · 24/02/2026

I’ve been experimenting with Spec-Driven Development lately, especially with Allium. I really like the simple approach it takes to keep the spec in sync with the code, and how naturally it fits into an adversarial testing loop juxt.github.io/allium/

juxt.github.io

Allium

An LLM-native language for sharpening intent alongside implementation. Velocity through clarity.

020

Antón @anton.galiglobal.com · 23/02/2026

I had the opportunity to work with OTAP Dataflow this week. It plays a role similar to parts of the OTel collector, but is built around Arrow’s columnar format with Rust components. I contributed support to persist telemetry to S3 in Parquet 👇 github.com/open-telemet...

github.com

Add aws s3 storage support to otap-dataflow by antonmry · Pull Request #2069 · open-telemetry/otel-arrow

Change Summary Add AWS S3 storage support to the otap-dataflow parquet exporter, mirroring the existing Azure storage integration. This includes configurable authentication methods (default credent...

020

Reposted by Antón

vectorware.com @vectorware.com · 17/02/2026

We are excited to announce that we can successfully use Rust's async/await on the GPU. This has never been done before. www.vectorware.com/blog/async-a... Supporting Rust's async/await (and futures) enables existing Rust code to work on the GPU and makes GPU programming more ergonomic.

vectorware.com

Async/await on the GPU

GPU code can now use Rust's async/await. We share the reasons why and what this unlocks for GPU programming.

525948

Antón @anton.galiglobal.com · 22/02/2026

Most code agents feel like web apps trapped in a terminal. I wanted the opposite. “Get React out of my terminal: a case for headless mode” is about composable CLI workflows, not unnecessary TUI bloat.

blog.anton.galiglobal.com

The anti-React Code Agent: stop shipping browsers inside my terminal

000

Antón @anton.galiglobal.com · 22/02/2026

TIL @grain.social, a photography-focused photo sharing platform using Atproto ❤️

011

Antón @anton.galiglobal.com · 19/02/2026

Made the jump from fzf to skim, and from pre-commit to prek. Rust tooling keeps surprising me with its quality. I gave nushell and helix a try, but they didn’t make it into my daily workflow. Nvim and Bash remain unchanged. zerobrew instead of homebrew is next in my list 🦀

010

Antón @anton.galiglobal.com · 18/02/2026

Surprising take in this paper: LLM-generated AGENTS md files hurt performance in most settings while increasing cost by 20%. arxiv.org/pdf/2602.11988

arxiv.org

142

Antón @anton.galiglobal.com · 17/02/2026

My colleague James shared some thoughts on how we work at Phoebe, including how we structure a large codebase so agents can operate more effectively. Looking at systems from an agent-first perspective is a powerful exercise. Good stuff in there www.linkedin.com/pulse/how-ou...

linkedin.com

How to Outship Teams 10x Your Size

The bottleneck in software engineering is no longer the typing. It's the intent: knowing what to build, specifying it precisely enough that an agent can execute it, and verifying that the result actua...

000

Antón @anton.galiglobal.com · 13/02/2026

Yesterday I received this message: “Antón, how’s everything going? I’m writing with some news I think you’ll like: Kynapses is dead. We decommissioned the cells this week.” I’m not really into bragging. Still, this project was special to me, easily one of the top five of my career.

110

Reposted by Antón

Antón @anton.galiglobal.com · 29/01/2026

Playwright or BrowserMcp aren't necessary anymore, Claude Code can connect to Chrome now and the setup is way easier code.claude.com/docs/en/chrome

code.claude.com

Use Claude Code with Chrome (beta) - Claude Code Docs

Connect Claude Code to your browser to test web apps, debug with console logs, and automate browser tasks.

111

Antón @anton.galiglobal.com · 22/01/2026

Some interesting takes in this article. I agree that code review needs to be rethink-ed in the era of code agents but we need more disruptive ideas cognition.ai/blog/devin-r... cognition.ai/blog/devin-r...

cognition.ai

Cognition | Devin Review: AI to Stop Slop

As code generation gets easier, code review is the new bottleneck. That's why we're launching a new way to quickly review and understand complex PRs in our latest tool for codebase understanding - aug...

031

Antón @anton.galiglobal.com · 15/01/2026

Testing opencode with LimaVM as a sandbox. Being able to run `opencode serve` in the VM and `opencode attach` from the host is very handy. Good news: it still works with my Claude Code subscription

010

Antón @anton.galiglobal.com · 13/01/2026

This paper documenting Claude Sonnet experiencing a complete meltdown is both fascinating and hilarious arxiv.org/pdf/2510.218...

arxiv.org

100

Antón @anton.galiglobal.com · 07/01/2026

Iván is a superb engineer. If you’re hiring, don’t miss this opportunity. You don’t get the chance to hire someone like him very often!

021

Antón @anton.galiglobal.com · 06/01/2026

"Coding Rust with Claude Code and Codex" matches my experience. Rust is still young, but coding agents are already solid on it thanks to the compiler. Bonus points for automating cargo clippy/cargo fmt and exposing bacon via tmux or zellij 🤓 dev.to/tigranbs/cod...

dev.to

Coding Rust with Claude Code and Codex

For a while now, I've been experimenting with AI coding tools and there's something fascinating...

041

Antón @anton.galiglobal.com · 28/12/2025

I don’t know what OpenAI changed this weekend, but ChatGPT and Cursor felt like a major downgrade. Even basic tasks failed, from image resizing scripts to Pokémon recognition. This is what makes working with big tech LLMs so frustrating: opaque updates and no changelog you can trust

000

Antón @anton.galiglobal.com · 26/12/2025

One last article on my blog to close the year. My work setup has changed a lot lately, even the way I interact with software, so here are my ramblings about it.

blog.anton.galiglobal.com

Git and Markdown are all you need

A minimalist personal stack: Git for ownership, Markdown for content, and tiny scripts plus AI to automate daily notes, news, and publishing

140

Antón @anton.galiglobal.com · 24/12/2025

Pro Tip: Press Ctrl+g in Claude Code to open the prompt in your editor. This feature will also be available in the upcoming version of Codex

000

Antón @anton.galiglobal.com · 19/12/2025

A love letter from @rmoff.net to Raycast is something you can't miss if you use a Mac. Raycast is the only tool I miss when using Linux and I learnt many good tricks reading the article rmoff.net/2025/12/18/a...

rmoff.net

A love letter to Raycast ❤️

110

Antón @anton.galiglobal.com · 16/12/2025

io_uring for High-Performance DBMSs: When and How to Use It arxiv.org/html/2512.04...

arxiv.org

io_uring for High-Performance DBMSs: When and How to Use It

000

Antón @anton.galiglobal.com · 13/12/2025

Superb article about building a queue in Rust 👇

051

Antón @anton.galiglobal.com · 06/12/2025

I once told a PM "We aren't interested in this feature, so let's skip the meeting." He insisted we meet anyway. He didn't want to sell me, he wanted my technical feedback. Brendan published a great post about this exact dynamic and how Intel is actually listening www.brendangregg.com/blog/2025-11...

brendangregg.com

Intel is listening, don't waste your shot

020

Antón @anton.galiglobal.com · 05/12/2025

Someone is rewriting the OTEL Collector in Rust, and honestly it’s exciting to watch. The Arrow ecosystem, otel-arrow in particular, is slowly taking over the OTEL tooling, one crate at a time www.youtube.com/watch?v=xeQn...

youtube.com

Introducing a Lightweight Rust OpenTelemetry Collector - Mike Heffner & Ray Jenkins, Streamfold

YouTube video by CNCF [Cloud Native Computing Foundation]

0111

Antón @anton.galiglobal.com · 27/11/2025

My zone 2 run was quite fun today

070

Antón @anton.galiglobal.com · 24/11/2025

I've discovered stacked panes in zellij (a rust version of tmux) this weekend and I can't live without them anymore

1121

Antón @anton.galiglobal.com · 21/11/2025

Spent way too long getting eBPF + Rust working on mixed setups (ARM Macs, Windows, Linux). Wrote up the approach that finally made everything click. Lima, Aya templates, uprobe examples, everything that actually worked.

blog.anton.galiglobal.com

eBPF with Rust using Aya (from macOS, Windows, or Linux)

140

Antón @anton.galiglobal.com · 21/11/2025

Sometimes life smiles back when you least expect it. I'm beyond happy to announce that I've joined Phoebe. I wasn’t planning to start this soon, but some opportunities simply cannot be missed 🚀 phoebe.ai

phoebe.ai

Phoebe - The immune system for your software

The immune system for your software. AI agents that continuously investigate live data, diagnose emerging issues and generate preemptive fixes.

050