Software estimation is hard – do it anyway

It’s well-established that estimating software projects is hard. One study by HBR found that one in six IT projects had cost overruns of over 200% and were late by almost 70%. Another study by McKinsey found that IT

Read full post on →

Latest News

July 2026

AI advice made people 3x less accurate but 2x confident, researchers found

Jul 20, 9:18 PM

Ana Maria Constantin writes how researchers found AI advice suppresses critical thinking. She highlights how people trust the wrong AI answers with |judgement suspension from 44% to 3%, accuracy from 27% to 9%, while confidence rose from 30% to 76%".

thenextweb.com

Hugging Face confirms breach affected internal datasets and credentials

Jul 20, 12:39 PM

Hugging Face disclosed a security breach where attackers exploited a vulnerability in a dataset uploaded to run malicious cod

techcrunch.com

The Kimi K3 Moment

Jul 19, 5:32 PM

Stephen Bochinski compares Kimi K3 (an open-source AI model) against Claude for coding tasks. He finds it's hard to tell the difference between K3 and Claude in terms of output quality and token efficiency, but at significantly lower cost ($3/$15 per million input/output tokens vs Claude's $10/$50).

stephen.bochinski.dev

Feature phones reimagined: HMD Nokia Dumb Phone

Jul 19, 9:16 AM

HMD (Human Mobile Devices) announces a new generation of smart feature phones launching through 2026, featuring a built-in digital wallet for financial access, video calling capabilities bridging feature phones and smartphones, and an AI as

hmd.com

What's the deal with all the random weekly quota resets for agents lately?

Jul 18, 7:06 PM

A developer reflects on the LLMs like Claude Code and Codex resetting weekly usage quotas mid-cycle, thereby giving giving users extra usage without paying more. While appreciating the free usage, he proposes that it's due to competition and business dynamics.

minimaxir.com

What Is an AI Agent Swarm and How These Systems Actually Coordinate

Jul 18, 5:40 AM

AI agent swarms are an emerging architecture where multiple specialized AI agents work together under an orchestrator to handle complex tasks. Unlike single monolithic AI models that handles everything end-to-end, swarms split tasks into subtasks.

startupfortune.com

China's open-weight Kimi model stuns AI world with frontier-level results

Jul 17, 12:51 PM

Chinese AI startup Moonshot AI has released Kimi K3, a massive open-weight AI model with 2.8 trillion parameters that appears to rival top American AI systems at a significantly lower cost.

axios.com

The Human-in-the-Loop is Tired

Jul 17, 12:21 AM

Pydantic co-founder explains how LLMs are creating a fundamentally destabilizing shift in developers' relationships with their work.

pydantic.dev

The Newest Way to Go Analog

Jul 9, 2:33 PM

The Atlantic explores a growing trend among Gen Z where they're deliberately adopting older, less connected digital devices (e.g. iPod Nano) as a way to escape the overwhelming choice and distractions of smartphones and streaming services.

theatlantic.com

The End of Reading Is Here

Jul 8, 12:08 PM

Optimists once believed that universal literacy was inevitable. Now it seems that the age of reading might be a short anomaly in human history.

theatlantic.com

Better Auth is joining Vercel

Jul 7, 3:52 PM

Better Auth is joining Vercel. We're joining to accelerate our work on open-source auth and securing agent workflows.

better-auth.com

The Making of Claude Code

Jul 6, 9:30 PM

The inside story of how Claude Code evolved from an internal command-line tool at Anthropic into today's full-fledged coding agent. The article features perspectives from researchers, engineers, and early users who contributed to its development.