Archive
Agentic UX ruins the coaching
AI coach
How to install a Claude Code skill from GitHub
Places to put the AI
Building a Google Slides renderer with coding agents
Search in program space
Three categories of guardrails
Week 1: the gap is inside your company
Core vs edge agents
Nobody gets fired for buying Anthropic
Your AI employees need a handbook
My agent use cases, part 2
My agent use cases, part 1
Possible
Agent skill - Statistics Estonia
Claude Code IS the software
Vibe coding a music learning app
Subtractive changes
Skills, not agents
AI Ranch podcast with Evalds Urtans
Principal
Anchor on what won't change
Invest into context
My deep research use cases
The complements to AI will increase in value
Shaping is high-leverage work for leaders
Are software jobs set to grow or shrink?
Gemini 2.5 Pro isn't quite Claude 3.7 at coding
Hiring as the marriage problem
Which European institution is running a massive LLM hackathon?
Will AI take us towards refinement of the self?
Half AI, half random curiosities
Vibe-code with stable infrastructure
Negotiating against an LLM
The GPU export limit might start to matter
Good LLM devtools are also good human devtools
LLMs highlight lack of progress in real world
Creating single-file apps with LLMs
Writing on mobile is different
DIY parametric wall art: plans to production
Best pastry in Tallinn
Frontier LLMs come every 1.5-2 years
Explaining quiet quitting
SAD lights
Strive for off-grid discipline
Momentum
Curiosity arises from lack of feed
Non-judgemental awareness is curative
Are LLMs deterministic?
Recent LLM launches, and LLM rumors
A wild speed-up from OpenAI Dev Day
LLM+API vs LMM+UI
Simplicity is essential in a generative world
Diverge-converge cycles in LLMs
LLM latency is linear in output token count
RAG is more than just embedding
Retrieval-augmented generation
Properties of a good memory
Things I've underestimated - Sep 2023
Launching OpenCopilot
Context engineering is information retrieval
Making GPT API responses faster
agentreader - simple web browsing for your Langchain agent
Why AutoGPT fails and how to fix it
Core innovations of AutoGPT
GPT-3.5 and GPT-4 response times
How data is used for LLM programming
Hacky multimodality
First thoughts on AI moratorium
Agents are self-altering algorithms
Index to reduce context limitations
Langchain frustration
Office 365 copilot is extremely impressive
Your non-AI moat
GPT-4, race, and applications
Alpaca is not as cheap as you think
Chain-of-thought reasoning
Define a vector space with words
AI product overhang
Self-driving apps
Text is the interface, not the use case
Prototyping with GPT API is fun
How do you decompose knowledge work for AI
AI shovels don't beat AI apps
Hands-on integrating NFTPort with Unity
ETH New York workshop: Launching your NFT app with NFTPort
Announcing NFTPort's $26M Series A
GPT and Me
20.1% yearly inflation in Estonia
HBS vs BAYC
Thoughts from my first web3 conference
Workshop: building NFT apps with NFTPort
Startup stock options: a beginner's guide
Web3: what I'm doing next
Foundering founder: the story of my startup year
Product before problem
Diffuse mode: thinking by relaxing
Tesla's data engine: the road to full self-driving
Becoming a definite optimist
If you can measure it, you can improve it
Video: Invest in Estonia, 1993
Integrate work and life
The negotiation technique that enables deep conversations
Stock options are hard
Commitment
Cause variance
Digital dopamine-seeking
School-career mashup
Courage in the face of variance
The office
To kill a mock-up, first
A productive creative
More pre-trained models, please
It's rocket science
Datasets: the source code of Software 2.0
Datasets carve the terrain of AI
Creating AI is Curating Examples
How do AI companies earn money?
Two meanings of "AI"
Your AI team needs DataOps
How to build your AI startup
Timid New World
ML at Veriff: What is possible in a year
Future of e-Estonia - a young engineer’s view
Data loops are the bottleneck in applied AI
Building automation-heavy products
The two loops of building algorithmic products
Why I Gave Up a 50% Higher Salary to Join a Startup
Units of automation
Writing a data science job description
Coping with the explosion of ML research
Side projects make the engineer
How to annoy a data scientist
What makes machine learning expensive?
Entering new fields from zero, depth-first
Making use of inclinations