Eduardo Romero

Software Engineer. Bringing Observability for financial data and business metrics to the innovation economy @ StandardMetrics.

Opinions are my own.

[PostHog @posthog]

https://t.co/zqNJaJFxZD

[Simon Späti 🏔️ @sspaeti]

Using DuckDB as a caching layer for sub-second analytics queries.

QuackStore is a DuckDB extension that caches remote files locally.
Block-based caching automatically stores frequently accessed portions.

https://t.co/DDusScC6FV

[ClaudeDevs @ClaudeDevs]

How do we automate business analytics with Claude?

New blog post covering our best practices for skills, data foundations, and evaluations when building agents to perform data analysis:

https://t.co/mfEJMAQFBU

[Samuel Colvin @samuelcolvin]

If only there was a runtime built specifically for this use case with strong isolation, repl support and snapshotting support! 🤷‍♂️

https://t.co/Rg4tNJJxgU

[Cloudflare @Cloudflare]

VoidZero, the team behind Vite, Vitest, Rolldown, Oxc, and Vite+, is joining Cloudflare. Vite stays open source, vendor-agnostic, and built for everyone. https://t.co/DJTpX4Q9Xt

[Simon Willison @simonw]

Uber reportedly now caps coding agents at $1,500/month per employee per tool - seems sensible to me, but it's also an interesting hint at the value Uber thinks these tools are providing
https://t.co/6YT0lCzPml

[Eduardo Romero @foxteck]

My agent asking me to get some R&R:

> When you're ready to open the PR, it's a self-contained review; the new experiment can be its own follow-up from the stash. Get some rest. 🌙

[Eduardo Romero @foxteck]

I've been writing / talking about the "Centaur Pod" teams a mix of Sr Eng + Mid/Jr Engs + Army of agents.

Bard just came up with two proposals for "Army of agents":
- A minion agent syndicate.
- A legion of minions.

hehe

[Anthropic @AnthropicAI]

Anthropic has confidentially submitted a draft S-1 registration statement to the Securities and Exchange Commission.

Pending completion of SEC review, this gives us the option to pursue an initial public offering.

Read more: https://t.co/onGZAhRLvD

[Eduardo Romero @foxteck]

This works pretty well in my experience (spec in the same repo where your code is).

It also has de added benefit of being a centralized place where you document how the it all changed as it was implemented (like ADRs).

Agents can just keep it up to date as they go.

[Eduardo Romero @foxteck]

It's lot better than the old Bard CLI, faster than Claude Code and smarter than Amp.

Also bullish. 🙌

[Eduardo Romero @foxteck]

agy (antigravity's CLI mode) is an odd duck:

> ● Schedule(Finished waiting 10 seconds.) (ctrl+o to collapse)

▾ Thought for 1s, 12 tokens
Awaiting Notification Completion

I've shifted my focus to a passive approach. No more active probing; I'm simply waiting for the notification. The tools have been set in motion. The ball is now in someone else's court.

My job now is merely anticipation.

[Eduardo Romero @foxteck]

Just realized that my "build a spec" ager guidance works just as well for reverse engineering a "system design" out of an existing code base 🤯

[Eduardo Romero @foxteck]

It also has some "personality" (like Claude's)... Funny IMHO. It resembles the mid journey/younger googlers I know xD

"Thank you for pair programming with me. Good luck with the improved pipeline, and let me know whenever you'd like to work on the next set of features!"

[Eduardo Romero @foxteck]

The new antigravity CLI (agy) sits between Amp and Claude in terms of speed and productivity.

It has its own internal task management with:
- plan
- tasks
- walkthru

I really like it, it mirrors how I work with my other agents.
Gemini's CLI was underwhelming.

[Eduardo Romero @foxteck]

The CLI itself has room from improvement.

That said, the harness (antigravity) and the model (Gemini 3.5 flash) are very impressive.

[Pydantic @pydantic]

"Monty runs cross-platform (no hypervisor or WASM backend), validates LLM-generated code against tool signatures with ty before any host tool fires, and supports Monty-native ResourceLimits for CPU / memory / output caps."

[Pydantic @pydantic]

𝗣𝘆𝗱𝗮𝗻𝘁𝗶𝗰 𝗔𝗜 𝘃𝟮 𝗯𝗲𝘁𝗮 𝗶𝘀 𝗵𝗲𝗿𝗲.
v2 is built harness-first. Much of what you configure on an agent flows through one primitive, the capability. One unit bundles tools, hooks, instructions, and model settings, so a whole extension plugs into every layer.

What's new 👇

[Matej Bukovinski @bukovinski]

Edit, redact, fill, annotate ... and VIEW PDFs directly inside Claude with Nutrient PDF Editor for Claude Cowork. Free preview. Install and use.
https://t.co/DvBUxkNP10

[Logan Kilpatrick @OfficialLoganK]

Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own.

We spent the last 6 months making sure Flash is great for real world use cases. It's available everywhere now!