Brandon

Committing to the bit every day.

[Brandon @xzendrex]

@ThePrimeagen Honest question, is 12k (~6k/w) *your* cognitive limit? You talk about how you enjoy actually reading/working on the code, so I’m curious how much of your tweet is satire vs genuine thoughts of burnout.

[Brandon @xzendrex]

Since we only had a bit of time w/ Fable, I almost exclusively used it to work on spec’s, skills, making dynamic workflows, etc, and I’m glad I did so.

Even though it’s gone, I’m luckily still able to feel its effects within my work, albeit less than before.

[shadcn @shadcn]

In light of what happened, I'm doubling down on skills like /improve.

A frontier model got pulled. If it happened once, it's gonna happen again. Fable today. 4.9 tomorrow or maybe gpt 6 one day.

So, treat intelligence as borrowed. Drain intelligence when it's available. Build a catalog of plans today. Then implement later with a cheaper, open source, or a model you control.

Build the backlog now.

https://t.co/rqHw0fPv4G

[Brandon @xzendrex]

@jarredsumner @ajcwebdev is his name sam?

[Brandon @xzendrex]

R.I.P. this guys wallet.

[Saïd Aitmbarek @SaidAitmbarek]

thinking of ditching codex for claude max 20x just for fable-5

is it worth it guys?

[Mountain Dew® @MountainDew]

Baja Blast is my one true OBSESSION

[Brandon @xzendrex]

@jarredsumner A single Claude run be pushing 58 commits a minute? No offense, but at that point, human-in-the-loop is more like human-gets-result-and-respins

[Sahil @sahill_og]

Me prompting Claude fable 5 at 3 am be like:

[Brandon @xzendrex]

I should have been more specific; Workflows can still skyrocket into millions if you let them. The env change just makes it so agents within them don't go beyond the base 200k or so.

[Brandon @xzendrex]

"CLAUDE_CODE_DISABLE_1M_CONTEXT": "1"

1M context windows are a trap. Stick to 200k or below. You will get drastically better results using smaller context windows and more narrowed/focused tasks.

This effects subagents and workflows as well.

[Brandon @xzendrex]

"My $200/m plan lasted 30 minutes on a single prompt, screw you Anthropic!"

The Prompt: "Use a workflow that spawns 300 sub agents to do XYZ because I don't know what im doing" ... 25 minutes and 10's of millions of tokens later on a single prompt... "limit reached".

[Brandon @xzendrex]

@MajPleasure Odds are they prompted an agent to make the prompt to another agent that delegated it 3 layers deeper before the workflow even began.

[Brandon @xzendrex]

"My $200/m plan lasted 30 minutes on a single prompt, screw you Anthropic!"

The Prompt: "Use a workflow that spawns 300 sub agents to do XYZ because I don't know what im doing" ... 25 minutes and 10's of millions of tokens later on a single prompt... "limit reached".

[BridgeMind @bridgemindai]

I hit my usage limits on my $200/month Claude Max subscription in less than 30 minutes using Claude Fable 5.

[Brandon @xzendrex]

Just another Monday morning where Cursor manages to uninstall itself once again.

@cursor_ai got famous for it's autocomplete. Now it's pivoting to auto-delete.

[Brandon @xzendrex]

@jarredsumner I’ve already made 4 reusable workflow scripts (I change core a lot in a monorepo), however the difficulties are in there being no API to reference unless you ask Claude to literally write down what it can do in a workflow, within a workflow…

[Brandon @xzendrex]

@jarredsumner Do workflow agents in CC have the codebase skills, rules, etc available to them or does that need to be specified in the workflow creation process.

Making some reusable ones, would like to know if I need to bake them in.

[Brandon @xzendrex]

Claude workflows are kinda nuts. It will be interesting to see how others try to replicate it within other harnesses. I very much expect Anthropic to release another plan soon, because first we had subagents and agent teams, now we have workflows. Tokens usage just going up.

[Brandon @xzendrex]

Seeing Jimmy mention he was a wiz head is amazing. If you see this @MrBeast, you should totally go back and play man. I get ur a busy guy, but the game really has evolved as we ourselves have grown up. I wouldn’t even do it for content, just enjoy the adventure like when we kids.

[MrBeast @MrBeast]

@CL207 For real, I loved Wizard 101 and WoW when I was a kid. I wish someone made an aged up modern version of them that I could get lost in.

[Brandon @xzendrex]

Glad that more people are seeing that the cost is our mental and cognitive load. I feel vindicated seeing .@shadcn with a similar mindset.

https://t.co/hjCvUNp0O9

[shadcn @shadcn]

Must read.

[Brandon @xzendrex]

Waiting for the @theo video talking about how public package registries are cooked and the future of code relies on releasing “prompts” for our own AI to implement changes rather than raw code or postinstall scripts.

[Socket @SocketSecurity]

🚨 Supply chain attack on the Laravel Lang organization:

700+ historical versions across multiple community-maintained Laravel Lang packages were compromised with an RCE backdoor, including:

laravel-lang/lang
laravel-lang/http-statuses
laravel-lang/attributes
Laravel-Lang/actions

The payload targets cloud creds, CI/CD secrets, Kubernetes tokens, Vault, browser data, password managers, SSH keys, and more.