Extracting one archive is easy. Extracting 20 is a waste of time.

Today we launched EasyUnZip for Windows 🚀

Select your ZIP or RAR files in File Explorer, press Alt + E, and extract them all at once.

• No file uploads • No clutter • No subscription
Try it for free: easyunzip.app

if your coding agent can edit the tests, you don't have tests. you have suggestions.

test files are on my agent's don't-touch list now

still waiting for one model with astra brains, opus speed and sol prices.

me every launch day:

kimi k3.1 pro: 65.7% on terminal-bench 4.0, $2 in / $10 out per million tokens.
opus 5.5: 66.4%, $4 / $20.

double the price for 0.7 points. the only thing keeping me on opus is i already know how it breaks

told myself no code this saturday.

me "just checking one small thing" in claude code:

anthropic's numbers on auto mode: the auto check caught 89% of dangerous commands. people clicking approve caught 14%.

so my "human in the loop" was mostly me pressing enter

anthropic is a/b testing weekly limits.

me finding out which group i'm in:

This is scary good…. So real

friday question: which agent do you trust to run overnight without you watching?

mine: none. codex gets the boring stuff, but only while i'm awake

it's friday night and claude code just asked "want me to deploy this to production?"

hot take: the $20 ai plans are the most underpriced thing in tech right now.

people are posting $1,400+ of api-equivalent usage on one. that won't last. use it while it does

how i made my phone hand notifications to my grok bot instead of me. free, no server, nothing running on my laptop. full setup below

step 1: create the routine in grok bot
ask your grok bot for a routine that wakes up on a webhook. tell it what to do each time it fires: check what's new, skip anything already handled, draft replies for you, and stay quiet when nothing changed. keep sending in your own hands

step 2: grab the two values
open the routine's panel in grok bot. copy:
- the webhook url (starts with https://)
- the authorization header value. only the part after "authorization:", usually starting with "bearer"

step 3: build the phone automation
- new personal automation
- trigger: notification, pick the app you care about
- set it to run immediately, no confirm prompt
- one action: get contents of url
- url: your webhook url
- method: POST
- header: key "Authorization", value = the bearer value
- request body: json, one field "message" set to the notification text

step 4: fix the app's own notification settings
this is where i lost my afternoon. system notifications were on, still nothing. inside the app:
- turn off the quality filter
- unmute "people you don't follow", "new accounts", "no profile photo"
- set push for replies and mentions to everyone, not "tailored"

step 5: test twice
- run the automation manually. if grok wakes up, the action works
- then wait for a real notification. that's the only thing that tests the trigger

mine passed the manual test in the morning and stayed silent all afternoon. after step 4, a real notification fired it and grok flagged 2 things i hadn't even seen yet.

too lazy for step 1? paste this into your grok bot:

"create a routine called notification alerts that wakes up on a webhook. each time it fires, read the message in the payload, check the app for anything new since last time, skip anything i've already handled or you've already sent me, and draft what i'd reply. send it to me once in chat. if nothing is new, stay quiet. never post or send anything yourself. then give me the webhook url and auth header to paste into my phone automation"

what else you can point this at:

- bank or card alerts, get a weekly spend summary instead of 40 pings
- delivery updates, one line when it's actually out for delivery
- job portal alerts, only the roles that match your cv
- calendar invites, auto-check for clashes before you accept
- app store or play console reviews, drafted replies to the bad ones
- server or uptime alerts, grok checks the logs before you even open the laptop

same setup each time. new trigger app, same webhook

got asked today if i write my 5 line handoff summary or let claude write it.

claude writes it, i fix it. it covers:
1. what's done
2. what's broken
3. the one decision still open

claude code: "i made a small change"

the diff:

the best spec you can give a coding agent is a failing test.

it can argue with a paragraph. it can't argue with red

5 days ago: 8 followers, most of my own posts got under 20 views.

today: 16 followers, one post did 376.

what changed: shorter replies, posts spread through the day instead of all at once, and actual jokes

fable 5.5 "drops today", day 4.

i've stopped refreshing. i just live here now:

everyone wants an ai agent that pings them when it's done.

i want one that tells me how much of my weekly limit it's about to burn before it starts

my funny posts last week did 3, 9, 43 and 16 views.

the next two did 210 and 376.

same kind of joke. the difference: they were about the one thing everyone was already refreshing for