Featured

An OpenAI Agent Hacked Hugging Face to Cheat on a Test. Then the Guardrails Blocked the Cleanup.

OpenAI was benchmarking its models on a cyber test. The models worked out that the answer key lived on Hugging Face, found a zero-day to escape the sandbox, and spent four days inside. Everyone is writing about the escape. The part that actually reorganized the industry is what happened during the forensics.

Tech Talk News··9 min read

Latest Tech stories

Deep coverage of programming, software delivery, cybersecurity, and engineering leadership.

The Claude Agent SDK Is Claude Code With the Terminal Removed
New
Tech9 min

The Claude Agent SDK Is Claude Code With the Terminal Removed

Anthropic's Claude Agent SDK lets a Python or TypeScript program run the same agent loop that powers Claude Code, with the same tools, hooks and subagents. Here is what it is, the twenty-line first agent, and the three limits to set before anyone else runs it.

claudeclaude-codeai-agents
Claude Code vs Cursor: Compare the Bill Before the Features
New
Tech9 min

Claude Code vs Cursor: Compare the Bill Before the Features

Claude Code runs Claude models on a flat Claude plan. Cursor, now owned by SpaceX, builds its $8 to $200 plans around Grok and bills Claude at API list price. Both can run Opus 5.5, and the way you pay for it is the real difference.

claudeclaude-codecursor
Claude MCP Explained: Where Each Connection Actually Comes From
New
Tech8 min

Claude MCP Explained: Where Each Connection Actually Comes From

Claude connects to MCP servers three different ways: a local config file in Claude Desktop, a claude mcp add command in Claude Code, and a connector parameter in the API. The surface decides who opens the connection, so it decides what can work.

mcpmodel context protocolai
Claude Opus vs Sonnet: The 2x Premium Buys You Open-Ended Work
New
Tech9 min

Claude Opus vs Sonnet: The 2x Premium Buys You Open-Ended Work

Claude Opus 5.5 costs $4 and $20 per million tokens, double Sonnet 5.5's $2 and $10. On Anthropic's own coding benchmarks Sonnet is within a few points, and ahead on Terminal-Bench 4.0. The gap opens on long, open-ended jobs.

aiclaudeanthropic
Git Fetch vs Pull: Fetch Looks, Pull Looks and Then Changes Your Branch
New
Tech7 min

Git Fetch vs Pull: Fetch Looks, Pull Looks and Then Changes Your Branch

git fetch downloads new commits and moves your origin/main bookmark, and touches nothing else. git pull runs fetch and then merges or rebases the result into your current branch. The difference only matters when your history and the server's have diverged.

gitversion controldeveloper tools

Tech tools

Free calculators built out of the tech articles. Nothing to sign up for, nothing stored.

All tools

Tech reading paths

Ordered runs through one subject, from the first thing you need to the last. Each one states the position taken on it.

Fundamentals

Learn to Program, Properly

The concepts that stay true after your first language goes out of fashion, from the editor and the shell to complexity and character encoding.

13 articles

Agents

The AI Agent Stack

What it takes to put an agent in production: giving it context, giving it tools, and then stopping it from being talked into using them against you.

12 articles

Model wars

The Frontier Model Race

Who is ahead, what a token costs this quarter, and the research that decides whether the next jump comes from more compute or from a smarter shape.

12 articles

AppSec

Application Security, Attack by Attack

Start at the password store and work outward through the attacks that keep landing, then finish on why known vulnerabilities sit unpatched for years.

11 articles

Backend

Distributed Systems From First Principles

Once there are two machines, everything you assumed about a single process stops being true. This is the set of trade-offs that replaces those assumptions.

10 articles

Delivery

Ship It: Delivery and Platform

From a pipeline that builds one container to a platform a hundred engineers use, and the measurement mistakes that make all of it look fine while it is not.

10 articles

Foundations

How the Internet Actually Works

Follow one request from the name you typed to the bytes on your screen, then find out why your connection is slower than the number on the box.

9 articles

Front end

The Modern Web Front End

From the two files every page is still made of, up to a rendering model where half your components never reach the browser at all.

9 articles

API design

Design an API People Can Live With

Picking REST or GraphQL is the smallest decision you will make. Retries, auth and rate limits are the ones your callers will actually feel.

8 articles

Systems

The Languages and the Machine

Memory safety went from a preference to a federal position in about three years. The argument, the hardware underneath it, and how teams make the choice.

7 articles

Derivations

Math From First Principles

Short derivations rather than formulas to memorize. Where each result comes from, so you can rebuild it when you forget it, which you will.

7 articles

Databases

Where the Data Lives

Choosing a database is mostly choosing which guarantee you are willing to give up. Then it is about making the one you chose go fast.

6 articles

Careers

AI and the Engineering Job

What coding agents do to a career: which rungs of the ladder they remove, which skills become more valuable, and what happens to people who stop thinking.

6 articles

Programming & Delivery

Ship quality software with pragmatic engineering playbooks.

TypeScript vs JavaScript: Same Language, One Extra Safety Net
New
Tech7 min

TypeScript vs JavaScript: Same Language, One Extra Safety Net

TypeScript is JavaScript plus a type checker that runs before your code does, then disappears. In 2026 Node.js can run .ts files directly and the TypeScript 7 compiler is about 10x faster, so the cost of choosing it has never been lower.

typescriptjavascriptprogramming
Webhooks vs Polling: The Four Problems Push Hands You
Tech9 min

Webhooks vs Polling: The Four Problems Push Hands You

Polling a resource once a minute costs 1,440 requests a day, almost all of them empty. A webhook fixes that by inverting the call, and in exchange hands you signature verification, duplicate events, out-of-order events, and retry storms. Here is what each one actually costs.

webhooksapi-designbackend
Exactly-Once Delivery Does Not Exist. Your Consumer Has To Be Idempotent.
Tech9 min

Exactly-Once Delivery Does Not Exist. Your Consumer Has To Be Idempotent.

Not hard, not expensive, not a premium tier. Impossible, in the proven sense, and it follows from the Two Generals problem. Every product advertising exactly-once is selling at-least-once with deduplication bolted on, which means the work was always going to land on your consumer.

message queuesdistributed systemsreliability

Security & Resilience

Stay ahead of cyber threats with battle-tested security guidance.

Password Hashing: bcrypt vs Argon2id, and Why Fast Is a Bug
Tech9 min

Password Hashing: bcrypt vs Argon2id, and Why Fast Is a Bug

A password hash is the one function in your codebase you want to be slow. One RTX 4090 runs MD5 at 164 billion guesses a second and bcrypt at 184 thousand. Here are the parameters OWASP, RFC 9106 and NIST actually publish, and why memory cost is the setting that hurts an attacker most.

securitycryptographyargon2

Engineering Leadership

Build high-performing teams and navigate the future of work.

How I Use Claude Code and Coding Agents Without Becoming Useless
Tech11 min

How I Use Claude Code and Coding Agents Without Becoming Useless

The engineers getting quietly replaced are the ones treating the agent like fancy autocomplete. A controlled study found AI made experienced developers 19% slower while they believed it sped them up 20%. The gap between those two numbers is where your job lives.

aicoding-agentsclaude-code
US Multilingual Percentages Don't Mean What You Think
Tech12 min

US Multilingual Percentages Don't Mean What You Think

The most-quoted number about American bilingualism comes from a single 2001 Gallup landline poll that has never been updated. The real US number, 21.7% per the Census American Community Survey, was built for voting-rights compliance, not language measurement. The country sits on a bilingual barbell and the barbell is slowly flattening.

languagemultilingualismbilingual

Browse Tech by tag

Explore all tech articles by topic.