391
Why LLMs Can't Write Q/Kdb+: Writing Code Right-to-Left
In my previous post I showed how LLMs struggle with coding in q/kdb+; having immediate feedback loops using the q-MCP server does help a…
392
A deep dive into self-improving AI and the Darwin-Gödel Machine
How the Darwin-Gödel Machine rewrites its own code to improve, and whether we can trust it
393
LLMs and coding agents are a security nightmare
Things are about to get wild
394
AI doesn't lighten the burden of mastery
<no value>
395
My 25-year adventure in AI and ML
This is the 25-year story of how I fell into the field of AI.
396
AGI is an engineering problem, not a model training problem
LLM models are plateauing, but true AGI isn't about scaling the next breakthrough model—it's about engineering the right context, memory, and workflow systems. AGI is fundamentally a systems engineering problem, not a model training problem.
397
LLM Inflation
One of the signal achievements of computing is data compression 1 : we take in data, make it smaller while retaining all information (“lossless” compression), transmit it, and then decompress it back to the original at the other end.
398
Salesforce regrets firing 4000 experienced staff and replacing them with AI
Salesforces has entered a phase of public reckoning after senior executives publicly admitted that the company overestimated AI’s readiness
399
AI Slop vs. OSS Security
griff Table of Contents Author's Note What Exactly is the Problem? Human Capital is Limited Burnout Burnout Burnout What AI Slop Looks Like The CV...
400
Fine-tuning LLMs is a waste of time
People think they can use Fine-Tune for Knowledge Injection. People are Wrong
401
If you are good at code review, you will be good at using AI agents
Using AI agents correctly is a process of reviewing code. If you’re good at reviewing code, you’ll be good at using tools like Claude Code, Codex, or the…
402
Building AI products in the probabilistic era
AI turns products from deterministic functions into probabilistic systems. That requires expanding old playbooks (SLOs, funnels, siloed finance), and reasoning in terms of trajectories, Minimum Viable Intelligence thresholds, and data as company operating system.
403
Reinforcement learning, explained with a minimum of math and jargon
To create reliable agents, AI companies had to go beyond predicting the next token.
404
Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book
New research could have big implications for copyright lawsuits against generative AI.
405
The behavior of LLMs in hiring decisions: Systemic biases in candidate selection
Hints of discrimination and lack of principled reasoning in frontier AI systems
406
My experience creating software with LLM coding agents – Part 2 (Tips)
moved to https://blog.efitz.net/posts/my-experience-creating-software-with-llm-coding-agents-part-2-tips/
407
The secret ingredients of word2vec (2016)
Word2vec is a pervasive tool for learning word embeddings. Its success, however, is mostly due to particular architecture choices. Transferring these choices to traditional distributional methods makes them competitive with popular word embedding methods.
408
Mini-R1: Reproduce DeepSeek R1 "Aha Moment"
Reproduce Deepseek R1 „aha moment“ and train an open model using reinforcement learning trying to teach it self-verification and search abilities all on its own to solve the Countdown Game.
409
Lack of intent is what makes reading LLM-generated text exhausting
There’s something icky about LLM-generated text when you think it’s written by a human. I think I finally put my finger on one reason why I feel this way.
410
Nvidia DGX Spark: great hardware, early days for the ecosystem
NVIDIA sent me a preview unit of their new DGX Spark desktop “AI supercomputer”. I’ve never had hardware to review before! You can consider this my first ever sponsored post …
411
412
About AI Evals
A guide to AI and LLM evals for engineers and product managers. Learn how to test AI systems, analyze failures, and improve retrieval, agents, and AI products.
413
Gemini figured out my nephew’s name
I wrote an MCP server to give LLMs read-only access to my emails. Here is the chat log.
414
Are polynomial features the root of all evil? (2024)
High-degree polynomial features aren’t inherently unstable: the power basis and unscaled inputs are. This post shows how regularization and Bernstein polynomials make polynomial regression behave.
415
Yes, Claude Code can decompile itself. Here's the source code
These LLMs are shockingly good at deobfuscation, transpilation and structure to structure conversions. I discovered this back around Christmas where I asked an LLM to make me an Haskell audio library by transpiling a rust implementation. An “oh fuck” moment in timeOver the Christmas break I’ve been critically looking
416
MCP vs. API Explained
Model Context Protocol (MCP) is an open standard that connects AI models to tools and data sources efficiently. This guide breaks down MCP’s architecture, benefits, and how it differs from traditional APIs
417
AI agent benchmarks are broken
Benchmarks are foundational to evaluating the strengths and limitations of AI systems, guiding both research and industry development.
418
The Future of MCPs
MCPs—Model Context Protocols—are set to transform AI from passive chatbots into powerful, action-taking agents. But the real story isn’t what MCPs enable—it’s who controls them.
419
Build your own Siri locally and on-device
That doesn’t spy on you
420
A visual exploration of vector embeddings
For Pycon 2025, I created a poster exploring vector embedding models, which you can download at full-size . In this post, I'll translate ...