J.G.Montoya

What are the things worth learning now as a software developer? What skills are still going to be relevant in the near, mid and long-term?

When the shorts get liquidated

Each step on an 'MCP chain' takes longer and consumes context.

CLI > MCP

How do devs in a team/company collaborate now in the age of AI?

This whole coldcard thing makes me really sad.

Seeing bitcoiners who really did nothing wrong get their life savings stolen hits hard.

I hope you can get back on your feet.

It's a harsh world. Being your own bank is hard. "Don't trust, verify" is hard.

If you used (enough) dice rolls you're safe.

The ColdCard bug:

Fixed-count loops are a bad definition of done.

loop: 3 can waste a pass after the work is already good, or stop with work still left. v0.13.0 adds until conditions to Shaka workflows. Set a maximum iteration budget, then stop as soon as a shell check passes or an LLM judge returns SATISFIED.

The LLM path is deliberately boring: Shaka only accepts SATISFIED when it is the final non-empty line. Every other result keeps the loop going.

The shipped review-and-fix workflow now works this way with a cap of 3. It reviews, fixes, and checks the review again until there are no issues that need code changes.

shaka run review-and-fix

v0.13.0 🤖 github.com/jgmontoya/shaka

I've been having an issue with claude code's CLI where it's stuck in an authentication loop.

/login "works" (it opens the site, I approve, then on the terminal it says "Login Succesful") but then it seems like the session/credentials are not really persisted, asking me to /login again.

Has anyone experienced this?

Am I the only one that's annoyed by all the retard takes of whatever-the-fuck being an "operating system" for X.

No. Your agents.md file is not an operaring system for AI.

When I'm optimizing some part of a codebase, I end up in the same loop. Try a thing. Benchmark. Keep what helps. Discard what doesn't. Repeat until the result is good enough, or until I get bored.

It works but it also eats half a day or more and ties up my attention while it runs.

AI doesn't get bored.

v0.11.0 ships the loop. Give Shaka an objective and it spawns an isolated git worktree, hands your terminal to your provider's TUI so a setup agent prepares everything with you watching, then runs the loop in the worktree until you stop it. Each accepted iteration commits, so the log is the record.

Inspired by Karpathy's autoresearch and davebcn87's pi extension.

shaka optimize start "<your objective>"

v0.11.0 🤖 github.com/jgmontoya/shaka

Recently I've seen a lot of people saying that Claude got dumber. Maybe it did. I don't know and it doesn't matter, because it reminded me why I don't want my setup locked to one provider.

Shaka has supported Claude Code and Opencode from the start. Skills, learnings, agents, commands, workflows, all portable between them. Define them once and use them interchangeably.

v0.10.0 adds even more optionality: Codex. Same skills, same learnings, same everything. Pick the best tool for the task and your setup comes with you.

Three new agents ship here too: CodeReviewer, DevOpsEngineer and TechnicalWriter. The slop scanner picked up some additional patterns with tighter false-positive rules to make your writing even less robotic and soulless.

v0.10.0
🤖 github.com/jgmontoya/shaka

Marmotize all the groups!

Shaka has had learnings since v0.3. The system extracts corrections, patterns, and preferences from your sessions and loads them into context next time.

The missing piece: what happens when you have hundreds of them. They compete for space. Related entries stay fragmented. Quality degrades under its own weight.

v0.8.0 adds automatic memory maintenance. Related learnings get condensed into compound entries. Cross-project patterns get promoted to global. Low-value entries get pruned when space is tight. Runs in the background after sessions, conservative by default, original learnings are archived before every change.

The system cleans up after itself now.

v0.8.0

🤖 github.com/jgmontoya/shaka

Last time I posted about Shaka it had slash commands. That was v0.4.0. Since then I've been building the parts that make each session better than the last.

It remembers across sessions now. Rolling summaries compress what you worked on into daily, weekly, and monthly context. You shouldn't have to explain what you did yesterday.

It runs multi-step workflows. Chain prompts, commands, and scripts in YAML. Steps pass output to each other. Add loop: 3 and it iterates: review, fix, review again until clean.

It catches AI slop. A WritingRules skill flags the "it's important to note" garbage, and Shaka scan scores any file so you can check before you post.

It has a skill manager now. Install from GitHub or the Clawhub registry. Skills get scanned for sketchy stuff before install.

And it ships a TDD skill that forces the AI into red-green-refactor instead of "write everything then pray."

v0.7.1.

🤖 github.com/jgmontoya/shaka

Also make it privacy focused for increased hardness!

Got media working on the

TUI

Inline image rendering, media upload/download and previews, profile pictures, reactions, message deletion, reply previews, NSec export, logout, follow-checks, view user profiles, create groups with members.

All done by directly using the White Noise CLI.

BTW you can point your agents to https://github.com/marmot-protocol/whitenoise-rs/tree/master/src/cli and tell them to try out the White Noise CLI. It works surprisingly well.

This was made to showcase what can be built with the new White Noise CLI, go check it out!

It was vibe-coded in a couple of hours, so beware that there might be some bugs with it, but it works! And the CLI is intuitive enough that AI can understand it and work with it pretty quickly!

It turns out, I completely fucked my follows list.

Nostr FTW!

White Noise CLI merged!!

Check it out: https://github.com/marmot-protocol/whitenoise-rs/tree/master/src/cli

wn messages send "This is how you will be able to send a message using the White Noise CLI"

Brewing some big plans 🦫

I created a small tool to search for stuff in nostr:

NST: Nostr Search Tool

It's very very basic but hey, if you happen to need a cli tool for searching stuff on nostr based on how close they're to you (or someone else) in the social graph you might find this useful

https://github.com/jgmontoya/nst

Each time someone adds nip-04 support to a client, a cute baby marmot dies.

I kept copy-pasting the same prompts between Claude Code and opencode. Custom instructions here, custom commands there. Two tools, two configs, same workflows duplicated.

Shaka v0.4.0 fixes that. /slash-commands, just markdown files with YAML frontmatter. Write it once, it compiles to both providers. Scope per project or go global. Override system commands with your own.

Ships with /code-review because that's one I was copy-pasting a bunch.

🤖 github.com/jgmontoya/shaka

I implemented a "user search" feature in White Noise, and it was slow as fuck.

So I had to do it again, properly this time.

I am pretty satisfied with the result. Next release of White Noise is going to ship with a pretty decent search feature.

You can read more about it here: https://blog.jgmontoya.com/2026/02/22/user-search.html

I feel like I'm programming more than ever, while also writing way less code than I used to.

Wild times.

A though just crossed my mind.

There might be one potential security benefit we might see from vibe-coding:

Now for very small functionalities we might end up using less dependencies, as the LLM would just code whatever is needed instead of looking for a package that does it.

The package probably does a bunch of other stuff you don't need as well, potentially opening up a bigger attack surface + 'supply chain' risk.

v0.3.2 released!

You can now see and manage what learnings has Shaka collected using shaka memory review

Pass it the --prune flag to get a suggestion of what to delete from Shaka itself

The reasoning framework got an upgrade as well!

github.com/jgmontoya/shaka

Part of intelligence is the ability to learn, adapt and improve.

Shaka just got smarter.

v0.3.0 introduces learnings: corrections, patterns, and preferences extracted automatically from sessions. Loaded into context at session start, filtered by where you're working so we don't bloat the context window, scored by recency and reinforcement.

This is a big step forward in the journey to building a truly intelligent assistant.

Windows support by

: he wanted to onboard a friend. Friend uses Windows (ugh, I know...). So he added the support and shipped the PR. Project's been open source for about a week.

Open source undefeated.

github.com/jgmontoya/shaka

Do you know if there's any good diving accounts on nostr to recommend to my good friend

who's new in nostr??

I use opencode and Claude Code. I don't want to pick a side.

PAI is a project that adds hooks, agents, skills and memory to AI coding assistants. I like a bunch of what it does. But the repo is hard to navigate, customization isn't obvious, and it only works with Claude Code.

So I did what any reasonable person does and built my own thing.

Shaka is my take. Clean directory structure you can actually reason about. Customizations that survive upgrades. It's still very simple and early stage, but I'm working on making it smarter.

It's been running my setup for a while now and it's been really useful to me, so I figured it might as well be useful to you!

Check out Shaka 🛠️ github.com/jgmontoya/shaka

What's the best free email service to use with openclaw?

I want to give my assistant its own email, what should I use?

What do you expect out of an AI personal assistant?

Big Plans: Flutter Edition

I paid for this hardware and I intend to use it

Kimi K2.5 released, with impressive bechmark results.

Has anyone used it for real tasks yet? What are your impressions?

https://x.com/i/status/2016024049869324599

What would the "real cost" of domething like claude code be if the model (with weights) was exfiltrated/public??

This but on nostr: https://shopify.engineering/ucp

Yesterday Chile rejected communism

https://world.hey.com/dhh/europe-is-weak-and-delusional-but-not-doomed-8b10e7cb

Working at the (ephemeral) WhiteNoise office https://image.nostr.build/af6abd0f7aa7b6fadcbc1ba563cb9001f80119e8267ad41ee81a9da39eb92313.jpg

We're cooking, and privacy is what's on the menu...

We gotta ship faster...