Colin

CEO @supernormal_app, building the last meeting assistant that does the work for you.

[Colin @ctres]

Workflows - we’ll hear that more than agentic soon.

[Colin @ctres]

@JosephDevoy Go go Joe!!

[Supernormal @supernormal_ai]

It’s now even easier to edit the slides Supernormal generates for you:

🖼️ Hover over images to swap them out

✏️ Click text to edit, add new paragraphs, or bulleted lists

📐 Adjust text alignment on hover

✨ Format text: bold, italic, underline, strikethrough

A polished post-meeting deck in minutes, not hours.

Get started with one of our slide templates here: https://t.co/j0IkB84EHe

[Colin @ctres]

@levie what we're seeing -> most RFP's are being drafted by the agency side after a meeting where the client mentions work that they want done. Turning that meeting into a shareable RFP, and response is a very common workflow now.

[Colin @ctres]

@_chenglou This is rad

[Colin @ctres]

@tobi yes @rousseaukazi !!

[Colin @ctres]

@martinengdahl @collision @stripe Glad to see someone got around to it!

[Colin @ctres]

@appenz @meetgranola @appenz give this a shot for free. And we'll take it one step further - you can give your agents/systems early access to our CLI: https://t.co/PtmBFInISf

[Colin @ctres]

@ideabrowser Free and has processed over 100M meetings: https://t.co/PtmBFInISf

[Colin @ctres]

@damienghader @cremedgtl @Lovable @Wktr Hit me up - we’ve got a project

[Colin @ctres]

@coffeewithone Let’s get you on Supernormal - supporting over 1 million organizations today.

[Colin @ctres]

@AlexandrePesant Swedish minimalism at its finest. Love it

[Colin @ctres]

@alexwikstrom Lfg!

[Colin @ctres]

DRY - don't repeat yourself - meets reality.

[BURKOV @burkov]

LLMs process text from left to right — each token can only look back at what came before it, never forward. This means that when you write a long prompt with context at the beginning and a question at the end, the model answers the question having "seen" the context, but the context tokens were generated without any awareness of what question was coming. This asymmetry is a basic structural property of how these models work.

The paper asks what happens if you just send the prompt twice in a row, so that every part of the input gets a second pass where it can attend to every other part. The answer is that accuracy goes up across seven different benchmarks and seven different models (from the Gemini, ChatGPT, Claude, and DeepSeek series of LLMs), with no increase in the length of the model's output and no meaningful increase in response time — because processing the input is done in parallel by the hardware anyway.

There are no new losses to compute, no finetuning, no clever prompt engineering beyond the repetition itself.

The gap between this technique and doing nothing is sometimes small, sometimes large (one model went from 21% to 97% on a task involving finding a name in a list). If you are thinking about how to get better results from these models without paying for longer outputs or slower responses, that's a fairly concrete and low-effort finding.

Read with AI tutor: https://t.co/MipHHO6rjX

Get the PDF: https://t.co/XQrqiaGwIO

[Colin @ctres]

This is the biggest shift in software since the cloud. We're building for it @supernormal_ai

[Colin @ctres]

See you for tomorrow morning’s ship

[Colin @ctres]

Stockholm morning ship.

[Colin @ctres]

@chamath This will be the saddest long term study that lacked when legislation was being crafted

[Colin @ctres]

@magnushambleton @WisprFlow @avec Decent

[Colin @ctres]

@bernhardsson Coffee this week - would love to talk sandboxes and have you meet @fffffffabian