RSSAmplifier

Blog

Blog by Grzegorz Kossakowski

AI ∩ Applications. Occasionally, sports, health.

gkk.devRSS feed ↗37 posts

Latest posts

Working with me

People often reach out to me for help or business conversations. Over time I’ve noticed a few recurring patterns that make it hard for me to respond well. This page is a quick reference for how to get the best out of an interaction with me.

Imperial mind trap

Ever since Russia attacked Ukraine, I’ve been trying to decode a nagging feeling about how the invasion is discussed in the wider world.

The rolling hackathon

Contrary to popular belief, Europe does not suffer from a capital shortage for funding technology startups. There’s ample early-stage money available—both in VC funds and angel networks.

Review of Tesla Model 3 in Poland: great car, but range anxiety is still on

Tesla Model 3 is an awesome vehicle but is severely constrained to being a commuter car due to the lack of Superchargers.

Europe: overfund the AI application layer

Europe should stop chasing “sovereign” LLMs and win by shipping vertical AI products where it already has distribution, domain depth or where the race is still at day one.

The charismatic prototype

Most of AI applications and the teams behind them face the challenge of the charismatic prototype - a phrase Shyam Sankar (Palantir's CTO) tossed in his recent interview.

Problem solving vs Problem picking

I frequently participate in debates about why Poles playing an outsized role in building frontier AI in the US doesn’t translate into an AI boom in their home country.

The AI Minimum Viable Team

The AI startup boom cycle comes with a reset: we're back to solving hard technical problems just to get products off the ground.

Come for the tool, stay for the network still relevant in 2024

I've been chatting with my friend about the origins of LinkedIn, recruiting through its social network a few others things.

Direct AI access is the 'command line' moment of app design

I came to believe that exposing direct access to an LLM or other AI is often an equivalent of command line access to a computer, with similar pain points.

2-4 years too early for AI in legal

I believe we're 2-4 years too early to apply AI in the legal space, and the bottleneck is not necessarily benchmark-measured LLM performance.

Prompt engineering tricks for LLMs

Most people overcomplicate prompt engineering. There are 100-page-long prompt cookbooks that can be distilled into a few techniques that have the highest leverage.

RAG without embeddings and AI infra companies

At Hotseat, we've hit a major milestone: the relevance of quotes from our PDF repository has jumped from 18% to 72%.

The anatomy of Hotseat's AI

Hotseat's anatomy - a few product design choices that take a raw and nascent technology of Large Language Models and turn it into a functional tool.

Towards high-caliber answers to AI-refined legal questions

I’ve been testing Hotseat - an AI-based Q&A for legal - with a small set of lawyers to see whether AI can produce high-caliber answers to legal questions.

You don't want to work with embeddings

Embeddings are the assembly of the AI world. We should spare people touching them and climb the abstraction ladder as fast as possible. As we did in software.

The OpenAI Keynote

The fate of GPT wrappers, increased experimentation, and AI developers.

Announcing Hotseat: AI in legal that's actually useful

I have yet to meet a lawyer who wishes they could spend more time at night solving a legal question.

AI-boosted development and specialization's wall collapse

In the new AI-boosted world, what excites me the most is that it allows me to get upskilled in unfamiliar specializations and take on more ambitious projects.

Escaping Gatsby, and moving to Eleventy

A rite of passage exists that switching JS libraries or frameworks is a form of procrastination.

The EU AI Act: an AI-generated changelog

You're an assistant at a law firm. Please summarise the amendments from June 14, 2023, applied to the proposed EU AI Act.

Generative AI is another driving seat

We’re in another cycle of AI moral panic. [...] The social fabric, social order, and the truth are about to be leveled by a savage AI bulldozer.

Tap Tap. Is this thing on?

Testing if publishing on my personal blog still works.

Bitcoin and software's plasticity

I've mostly remained oblivious to the crypto movement to date. I didn't have a strong opinion on the tech in either bear or bull direction.

Which Way Do You Run?, part one (aka part five)

Which Way Do You Run?, part one (aka part five)

The slimshady year, part four

Sell it like you mean it

The Slim Shady Year, Part Three

Diagnosing the muscle's operating system.

The Slim Shady Year, Part Two

Pilgrimage to the (in?)famous chiropractor

The Slim Shady Year, Part One

Undoubtedly, my recovery from a rare and daunting condition is the best story I can tell.

Data science in biology

As a study of life's fundamentals, microscale biology is going through a paradigm shift. Data science is at the center of it.

The Great Influenza - book notes

When COVID-19 struck, I picked up a book on one the most impactful pandemic in world's history to understand what we're to face in the coming months. I was trying to work out the mechanics of a pandemic, if you will.

The status of Kentucky Mule (faster Scala compiler)

Occasionally, someone on the internet comes across Kentucky Mule and I receive an email asking its status.

Quickstart Molecular Biology

A crash course in molecular biology on 160 pages if you need one.

The right metric to track COVID-19 in Europe

Every multi-dimensional crisis requires tight feedback loops for effective assessment and management; COVID-19 is no exception.

What I know about COVID-19

Like many people, I’ve been glued to the news over COVID-19. In many ways a perfect distraction: it’s an endless source of news, there’s a lot of inherent uncertainty and drama and it’s full of numbers.

Drug repurposing with an algorithm and COVID-19

Programmers will recognize in drug repurposing a hack. A notably rewarding one.

Hello, world!

I'm experimenting with publishing blog posts here instead of on Medium.