Currently reading (Show yours?)

who TF approves this shit. i mean what can go wrong.. ``` Next 16.4 includes an option, experimental.agentUpgrade, that defaults to 'security'. Before each build, Next checks the installed version against published security advisories. If the version is affected and you're running in an interactive terminal (and not in CI), it shows a menu offering to hand an upgrade prompt to a coding agent it finds on your machine (lib/upgrade/harness.js, called from build/index.js:545). ```

chrome has no built-in way to import history from another chrome install. this script copies chrome history to a local profile without overwriting. will keep adding my chrome transport utilities here. github.com/rnjn/chronoport

this is such a fascinating read anthropic.com/engineering/re… so much to learn...

much fun building this. more while testing. coming soon from @base14io !

herdr has been such an amazing tool for me for the last 2 months or so, its made a huge difference to my flow. thank you @herdrdev

animejs is so cool

As a PM, the first pass is yours. If you are still waiting for someone else to do it, you are the bottleneck. A PM's week used to run on requests. Someone pulled data. A designer drew the mock-up. An engineer built the prototype. Every request joined a queue, and the every decision waited for clearance. That reason to wait is gone. If you are a PM today, these are the basic expectations: 1. You pull your own data. AI writes the query. You know your data well enough to catch a wrong number. 2. You read customers at volume. You have AI label thousands of tickets, one row each, with a verbatim quote, and you check a sample against the source. 3. You sketch the basic UX. You describe the flow precisely, including the states people forget. 4. You build the first prototype. You know what it can show (people understand the idea and can use it) and what it cannot. 5. You review with AI before the team sees it. You ask for the strongest objection from a customer, an engineer, a support agent and an executive. This is the baseline now. A PM who skips the first pass is choosing to wait, and the decision waits with them. They are as good as dead weight for the team. Look at last week. What did you wait for that you could have started yourself?

We bet on OpenTelemetry-native from day one because this was always where things were heading. Now FastAPI ships it built in. Instrumentation is becoming a framework default, not a chore. Upgrade to fastapi[standard] 0.142.2 and send those spans our way at base14.io 🔭

We've added the Azure UK region to our fleet. If you need lightning-fast unified observability (logs, metrics, traces, prompt management, evals) and are hosted in this Azure region, please reach out. (we serve 14 regions across AWS/GCP/Azure/Hz and growing!)

been cooking for a while, Scout Assist coming soon 👇

Being a founder is mostly discovering that every trolley problem leads to approximately three more.

are these the most annoying mandatory selectors ever made by a human being ?

tests are context. write them first.

communicating with people's agents is so much easier than communicating with them

I’m not kidding this is such a pressure release. Finally I’m not talking to a self obsessed midlife crisis wala uncle who’s obsessed about explaining every punctuation mark.

always put your retries behind feature flags. thank me later githubstatus.com/incidents/z…