Doneyli De Jesus

There should have been provisions in that investments that required it to be fully investable in Canadian assets and interests. If that wasn't the case, that's on the Gov. Like you said, Cohere as a company will do what's best for their company and its investors etc. The discussion on if Canada even has the capacitiy, skillset, etc. to deploy these AI data centers is a whole different discussion.

Please. Don’t go after the company. They have a responsibility to make business decisions that are sound for the company. Not a government program. We need some sort of “this is how global tech works” for MPs.

Modern day FOMO! When you are about to “leave tokens on the table” and have other life commitments 😭 I need to ask my agents to do more work before tomorrow. Who else is struggling like this? 😅

@Cappy_Nate @Cappy_Nate I had to create a passcode but I should be on the be X chat

@doneyli Cool! Gives me this. X is annoying.

Started an Actual Canadian Builders group chat. -Hardware -Software -Space -Defence -Web3 -Health -AI -Energy -Manufacturing You name it. But it's a Canadian only zone. 🇨🇦 If you want in, drop your name in the comments and ping anyone who should be there! LFG! 💪🇨🇦🏗️🚀

@Cappy_Nate Uhm, let me check! Thanks for adding me

o11y for the win! I built something similar a few months back. It's leverages a stop hook and consolidates in Langfuse. I have all my Claude Code sessions from my laptop, my server, etc., pointing to the same instance. It works amazing! Keep up the good work. github.com/doneyli/claude…

Claude Code doesn't show you how many tokens you're using for subscriptions. No breakdown by model. No breakdown by project. Just a progress bar that says "63% used." So I built a local dashboard that reads the files Claude Code already writes to your machine. Turns out every session, every turn, every token is logged to ~/.claude/projects/ in JSONL files. Input tokens, output tokens, cache reads, cache creation, model name, timestamp. It's all there. You just can't see it. My numbers over the last 30 days: 440 sessions. 18,000 turns. $1,588 in API-equivalent costs. On one day, the cache spiked to 700M tokens - visible cache bug, two days in a row. The dashboard scans those local files, builds a SQLite database, and serves charts on localhost:8080. Filter by model (Opus, Sonnet, Haiku). Filter by time range (7d, 30d, 90d, all time). Cost estimates based on current Anthropic API pricing. Works retroactively. First run processes your entire Claude Code history. Install: git clone github.com/phuryn/claude-… cd claude-usage python3 cli.py dashboard Windows: use python instead of python3. Zero dependencies. Python standard library only. Open source, MIT. Star it. Fork it. Make it your own.

@fluffleworld Incredible. That conversation data has a ton of value and glad to see you are creating a community and ecosystem around it.

Anthropic just cut third-party harnesses off Max subs. Makes sense, running agents 24/7 on a flat rate isn't sustainable. But the users affected are the exact ones who know how to swap models. Memories are markdown files. Brains are replaceable.

Full post (free preview + paid implementation with 18-skill template repo): doneyli.substack.com/p/build-your-a… Subscribe for weekly builds + frameworks: doneyli.substack.com/subscribe

Claude you work for me. Not the other way around!

It turns out, most applications of AI can be solved with Code… which AI has gotten extremely good at. Even non-tech use cases, benefit from the massive coding capabilities. In retrospect, Anthropic’s decision to go all in the coding use case has paid off massively.

"The purpose of your job is not the same as the tasks of your job." Been in enterprise conversations for 3 months where that one sentence explains everything people are feeling but can't articulate. Built a framework around it. New newsletter just dropped. doneyli.substack.com/p/the-knowledg…

Vibe coding has officially hit mainstream. My honest take: yes, get the account. Build an AI twin you can rap battle with. That would actually be incredible. But I hope creatives don't lean too heavily on AI. There's something je ne sais quoi about knowing a human wrote that bar, painted that canvas, composed that melody. We can use AI to amplify creativity. We shouldn't use it to replace the thing that makes creative work matter.

I need a GitHub too! Is it like that or nah?

I've been using Claude Code + cmux to run my Night Shift agents and it has been amazing. Now my workflow every night is: 1. Plan the work for the night. 2. Create discrete gh issues 3. Invoke my /night-shift skill pointing to the gh issues That's it! Every agent has its own worktree that I can review in the morning. It will save progress and update memory as they go. Here's what I include in my skill: # 1. Create a surface cmux new-surface --type terminal --workspace $CMUX_WORKSPACE_ID # 2. Launch claude with worktree + system prompt cmux send --surface surface:{ID} --workspace $CMUX_WORKSPACE_ID \ "claude --worktree \"issue-{N}\" --model {model} --permission-mode auto --effort high \ --append-system-prompt-file {PROJECT_ROOT}/.night-shift/{DATE}/agents/{N}/system-prompt.md" cmux send-key --surface surface:{ID} --workspace $CMUX_WORKSPACE_ID Enter # 3. Wait for claude to load, then send the prompt sleep 5 cmux send --surface surface:{ID} --workspace $CMUX_WORKSPACE_ID \ "You are night-shift agent for issue #{N}: {title}. Write all working docs to .night-shift/ in your worktree root. Begin with Phase 1: Understand. Full issue body: {body}" cmux send-key --surface surface:{ID} --workspace $CMUX_WORKSPACE_ID Enter # 4. Rename the tab cmux rename-tab --surface surface:{ID} --workspace $CMUX_WORKSPACE_ID "NS: #{N} {title}"

@MachineMktgShai So I have Gatus monitoring the launchd jobs on my MacOS. I instrumented in a way that detects if the agent failed, it's running, etc.

This is exactly right. The overnight agents are where it gets powerful. While you sleep: - Content being drafted - Signals being tracked - Research compiling We're running a similar setup. The compounding effect is real. Curious about your signal tracking approach. What sources?

Building a content marketing system for solo founders and solopreneurs with Claude Code doneyli.substack.com/p/how-i-automa… This goes beyond just automating copy. It has night shift agents tracking signals, leads, opportunities while you sleep. More time for founders to work ON the business

@nblasgen Yeah I was doing the same. It has been happening more frequently this week during peak times. In particular around 5pm EST.

@doneyli I switched to Sonnet (1m) and it's working now. Clearly not as good as Opus, but beggers can't be choosers I guess

Go out and get some fresh air. Claude is Overloaded 😤