dayson — e/llm

Let's FAFO!
AI Acceleration (@Google / Gemini Android) by day. TUI-Pilled Python by night. Previously 2X startup.
(opinions are my own)

[dayson — e/llm @dayson]

@Velocita_Prime Gooooo team!!!

[dayson — e/llm @dayson]

@thdxr really miss that OPENCODE_EXPERIMENTAL_SESSION_SWITCHER with the preview column you guys had going there for a while! did it get punted in latest release? 😢 @opencode

[Peter Steinberger 🦞 @steipete]

Here’s your monthly reminder that you shouldn’t be prompting coding agents anymore.

You should be designing loops that prompt your agents.

[dayson — e/llm @dayson]

@thsottiaux Easier devcontainer support for codex cli, that just works out of the box, the sandbox bubblewrap foo, can almost never get the right config combination between host and devcontainer. Not an issue with other coding agents.

[dax @thdxr]

i have seen enough proof now that using a coding agent is a deep skill

it's confusing because the people you see heavily using them produce horrible results

but that's because it's a skill! you can get better and the ceiling seems pretty high - this is very exciting to me

[dayson — e/llm @dayson]

honestly, this is one of the absolute coolest and most practical on-device AI features to ever hit a smartphone. I feel incredibly lucky to have played a part in bringing it to life.

[Sameer Samat @ssamat]

Scanning documents from a phone is a pain!

Glad to see the new document scanning experience in Google Drive on Android is rolling out now.

📄 Smart Batch Scanning: Scan multiple pages at once, automatically splits them into separate docs.
🚫 Duplicate Detection: Hovering over a page you already scanned? The system instantly catches it.
📸 Auto-Best Frame: We automatically swap out blurry scans for the highest-quality frame available.
🔒 100% On-Device Processing: Lightning-fast performance, offline availability, and total data privacy.

Rolling out now on Android; Works on devices with 8GB+ of memory.

[dayson — e/llm @dayson]

@MrNigelNg oh man honestly awful food. Anything in China town is baseline tastier than the 4 dishes we tried here :( even Jamie would not approve!

[dayson — e/llm @dayson]

@thdxr @ucirello jj support would be killer!

[Suhail @Suhail]

The pace of frontier AI progress is moving so fast that you could spend an entire day on X reading every interesting thing people publish and get absolutely nothing done. Somehow you have put the blinders on and build.

[dayson — e/llm @dayson]

- would be sleek to just have opencode attach the voice file and send it over to models that can natively support it!

- painpoint in all other TUIs - their /voice and image paste features never work when connected over remote-ssh or in devcontainers -- thoughts on a clean solve for that?

[dayson — e/llm @dayson]

"The company moves at the speed of its slowest secret"

[tobi lutke @tobi]

https://t.co/5MJ7u1VHwf

[dax @thdxr]

finally found the right metaphor for this shift in how i use opencode.

i used to treat it like 3D printing, where you build the thing layer by layer and commit to each piece as you go

now it feels more like progressive rendering, you start with a blurry version of the whole thing, then keep making full passes over it, and each pass sharpens the entire shape

doing this with gpt 5.5 and voice prompting is the first time things feel like they're clicking

[dayson — e/llm @dayson]

@dhh @OpenAI Are you using it with opencode these days or via codex TUI directly?

[dayson — e/llm @dayson]

"Models are already so good, you need to raise your ambitions or you won't realize just how good they've become. If you don't raise your ambitions, you'll think they've stagnated..."

[Simon Smith @_simonsmith]

Riley's recent tests and various posts about GPT-5.5 have revealed something about model progress. Models are already so good, you need to raise your ambitions or you won't realize just how good they've become. If you don't raise your ambitions, you'll think they've stagnated. Riley keeps doing incredibly ambitious stuff and GPT-5.5 delivers.

[Jason @mytechceoo]

CEO obsessed with token maxxing

[JNS @_devJNS]

me and Claude building an app at 3AM

[dayson — e/llm @dayson]

vibe-pager, that llm artifact before your 1-pager and 2-pager

[dayson — e/llm @dayson]

Yacht Problem
/yät ˈpräbləm/

[Noun]
A concern that seem might seem pressing
to a junior engineer, but you know can wait until after IPO

intern: "There's an unfinished TODO in this PR!"
senior: "...that's a Yacht Problem"

https://t.co/1BLNh11fvd

[Om Patel @om_patel5]

SOMEONE TURNED THE VIRAL "TEACH CLAUDE TO TALK LIKE A CAVEMAN TO SAVE TOKENS" STRATEGY INTO AN ACTUAL CLAUDE CODE SKILL

one-line install and it cuts ~75% of tokens while keeping full technical accuracy

they even benchmarked it with real token counts from the API:

> explain React re-render bug: 1180 tokens → 159 tokens (87% saved)
> fix auth middleware: 704 → 121 (83% saved)
> set up PostgreSQL connection pool: 2347 → 380 (84% saved)
> implement React error boundary: 3454 → 456 (87% saved)
> debug PostgreSQL race condition: 1200 → 232 (81% saved)

average across 10 tasks: 65% savings. range is 22-87% depending on the task.

three intensity levels:

> lite: drops filler, keeps grammar. professional but no fluff
> full: drops articles, fragments, full grunt mode
> ultra: maximum compression. telegraphic. abbreviates everything

works as a skill for Claude Code and a plugin for Codex.

this is PEAK