abhinav.co

cto • probabilistic at the edges, deterministic at the joints • ex nintee, urban company, hike, yahoo • bits pilani

[José Valim @josevalim]

Elixir v1.20 released! Now officially a gradually typed language: Elixir type checks every single line of code, finding bugs and dead code, without developer overhead (no typing signatures) and extremely low false positives rate. Plus a faster compiler! Links and reports below.

[abhinav.co @abhinav]

Most of the Indian tech teams have folks in two extremes - one using AI tools sub-optimally and one that is over reliant on these tools.

I am not sure what to make of this though.

[abhinav.co @abhinav]

https://t.co/biXDPHHxgn

[IndiaToday @IndiaToday]

A massive dust storm swept across Rajasthan’s Churu on Saturday afternoon, turning day into night as thick clouds of dust engulfed the sky. Visibility dropped drastically, forcing motorists to switch on headlights and slowing traffic across the region.

#Churu #Rajasthan #DustStorm #Sandstorm #WeatherUpdate #IndiaWeather #indiatodaysocial

[abhinav.co @abhinav]

Extremely severe alert

[abhinav.co @abhinav]

My Wi-fi was named Gulzar, and now it's named Katseye. I don't even know when this transition happened!

[abhinav.co @abhinav]

TIL: "ig" that I keep seeing on chats mean "I guess"

[Ilya Grigorik @igrigorik]

AI made protecting maker time 10x more important.

Maker time was always about protecting wetware context. What changed is the magnitude: we can now take on problems with *much* larger scope, complexity, and moving parts. The context window has to carry more problem state, more connections, and more nuance than ever before.

A "quick sync" doesn’t just take 15 minutes; it evicts the cache... and our wetware has no instant snapshot restore; unlike silicon, we refill context at glacial speed.

Protect your 🧠 wetware cache / context window!

[abhinav.co @abhinav]

4/

Putting the schema in the file gives the agent the same context a human would otherwise have to go dig for.

Writeup + tool (MIT licensed):
https://t.co/8yKrnCWkjt

[abhinav.co @abhinav]

1/

In the Rails world there's a small gem called annotate_models.

It writes your database schema into the top of each model file as a comment: every column, its type, what's nullable, which columns are indexed.

Nothing fancy. But you never have to leave the model.

[abhinav.co @abhinav]

3/

The schema already exists in your system. This just puts it where the reader is actually looking.

And increasingly that reader is a coding agent: it sees the model, not the migrations or the live database.

[abhinav.co @abhinav]

Now reading

[abhinav.co @abhinav]

@captn3m0 Oh great, let me check it out in the evening. Also will checkout the actual website too - it's a good one.

(Interesting to see you are using sqlite with Jekyll)

[abhinav.co @abhinav]

Made my blog agent-friendly and wrote two small Jekyll gems while doing it. Both on RubyGems now.

- jekyll-markdown-output: a clean .md sibling for every page.
- jekyll-llms-output: /llms.txt + /llms-full.txt.

(1/5)

[abhinav.co @abhinav]

MIT licensed, on RubyGems, working live on my site:

- https://t.co/K3I0Tbv8Xe
- https://t.co/CY9vqsL3XD

Gems
- https://t.co/zkjHSvpBEW
- https://t.co/PLNLZvHunz

If you, like me, are still using Jekyll, give both the plugins a try

(5/5)

[abhinav.co @abhinav]

Why both?

The .md siblings give an agent a clean fetch when it lands on a page.

/llms.txt + /llms-full.txt give it a way to discover and ingest the whole site without crawling.

Different problems, separate gems.

(4/5)

[abhinav.co @abhinav]

@HashNuke Thanks, Akash. Inspired by Oban primarily.