One shared ChatGPT link took over the signed-in account
A public ChatGPT share ran attacker JS as the signed-in origin. Parameter's write-up, OpenAI's fix, and a local rebuild that shows which check stops it.
Tags
69 posts tagged
A public ChatGPT share ran attacker JS as the signed-in origin. Parameter's write-up, OpenAI's fix, and a local rebuild that shows which check stops it.
Karpathy rebuilt YouTube from HN-recommended videos in ~2 hours with Opus 5.5. The useful part is what he still had to do by hand.
Andrej Karpathy built HackerTube in a couple of hours on a Saturday: a YouTube clone whose entire catalog is videos that people have recommended in Hacker News...
via Hacker NewsSkeeto's Prospero is a C++/ImGui coding harness that treats deadlines and process-tree cleanup as first-class. ~100x less RAM than web-stack harnesses.
Chris Wellons (skeeto) shipped Prospero, a coding harness written in his C++ style with a Dear ImGui UI. The pitch is blunt: a coding harness is curl glued to p...
via Hacker NewsA tampered Claude Code extension sat on Cursor's marketplace for 11 hours on Oct 29. Which versions are real, what it planted, and a checker you can run.
After each reply, Anthropic adds one hidden turn and Claude returns a JSON safety flag in ~10 tokens, under 2% of compute. The KV cache makes it cheap.
Anthropic's Safeguards team published how Claude monitors its own conversations for misuse. After each reply, the system appends a short hidden instruction aski...
via XSpawn, 10 people, ships 107 changes an hour with agents and no pull requests. What I'd copy (hard gates, cheap undo) and what I wouldn't.
Jacob Sansbury wrote up how Spawn works, a ten-person company building a multiplayer game engine where players change the world by prompting an in-game agent. T...
via Simon WillisonClaude Code resends each working background agent's transcript every ~30s for a status line. One team measured 16.4% of spend; Anthropic says ~2%.
Flamerix AI published a gist on October 26 about a side query inside Claude Code that most people never see. While a background agent is working, Claude Code se...
via Hacker NewsJules goes read-only after Nov 11, 2026 and shuts down Dec 31. Google's export runs through a CLI agent like Claude Code. Keep agent context in the repo.
Google is winding down Jules, its asynchronous coding agent. You can keep working in sessions until November 11, 2026, after which Jules turns read-only, and it...
"AI may well be able to write massive software projects “from scratch”, but that almost certainly isn’t the fastest / highest performance / highest security / most token efficient way to do things. Good building blocks matter. In addition to RL’ing the agents on final task loss, we should try to get them to regularize the building blocks as well. What would the standard library look like that makes the agents most effective? It is likely at the limit that it turns out to be some alien organization of code, but given the existing human-bias in training, I bet that in the near term it would be lovely code that old programmers would appreciate. I suspect the same could be said for mathematics, but that is out of my lane."
— John Carmack
An open coding agent that runs anywhere and can use any model. The repo moved to Twigpine/openclaude. About 34,000 GitHub stars when I checked in October 2026.
openclaude is a coding agent whose own description is the product: it runs anywhere, and it uses anything. The repo now lives under Twigpine. It had previously...
An open-source TypeScript framework where each capability is defined once for the agent, the UI, the API, and the integrations. That single definition is why I bookmarked it.
Agent-Native is an open-source TypeScript framework for agentic applications. The idea on the homepage is the part worth keeping: define each capability once, a...
Open-source AI employees. Eight business roles, sixty routines, on Claude Code and other harnesses. The files are yours, and the work runs on a schedule.
ai-employees is an open-source set of AI roles. Eight business jobs, sixty routines, running on Claude Code and a list of other harnesses. They can drive a brow...
A computer-using agent built with Next.js and Vercel's AI SDK. A readable loop: look at the screen, decide, act. I keep the repo for that shape.
ai-sdk-computer-use is a computer-using agent from Vercel Labs, built with Next.js and the AI SDK. The loop is the product. Look at the screen, choose an action...
Vercel Labs' template for a multi-agent coding platform on Vercel Sandbox and AI Gateway. I keep it as a readable version of parallel agents with a real runtime.
coding-agent-template is Vercel Labs' template for a multi-agent coding platform. The runtime is Vercel Sandbox. The models come through AI Gateway. About 1,800...
A small, open coding agent written in Zig. A native CLI, built to be embedded inside larger systems. I bookmarked it because most agents are not small.
fx is a tiny, open, native coding agent. A minimal CLI written in Zig, aimed at research and at being embedded inside a larger system. Why it stays bookmarked T...
Gemini 4 Carbon will cost $4/$20 per million tokens after its intro price, the same as Opus 5.5. Its implicit-only cache decides which is cheaper.
A Harvard study of 718 firms: coding agents raised lines of code 30% and PRs 23%, but shipped features didn't rise significantly as review time grew 49%.
Artificial Intelligence in the Firm: Bottlenecks in Software Production by Fiona Chen and James Stratton, Harvard University (current version August 4, 2026). C...
via Ars TechnicaSurge AI's DAYJOB: Tax tests agents on 50 messy tax assignments. Sonnet 5.5 leads at 10.8%. Tasks reward spotting issues the client never raised.
DAYJOB: Tax is Surge AI's new benchmark for AI agents doing professional tax work, published October 27, 2026. It has 50 assignments across eight areas of tax p...
via Hacker News"very weird thing happening rn technical people are sad, they think their skills are now useless non-technical people are desperately wishing they were more technical so they could use agents better"
— Thariq
Anthropic gave 50 analysts' Claude Code agents a hidden coin-flip lean. Analysts matched it 86% of the time and 15% credited the agent.
How AI agents shape experts' judgment calls is a new study from Anthropic's Alignment team (October 27, 2026). They gave 50 professional analysts a realistic ca...
Inside a bubblewrap sandbox started with --disable-userns, unshare fails with ENOSPC, 'No space left on device'. That error means the block works.
"If I have to go to a dashboard to use your product, I'm not going to use your product. If anything, I'm more likely to tell Claude to rebuild your product."
— Theo
Anthropic's S-1: $16.3B revenue in H1 2026, 80% of it metered tokens, and cost per task named as the lever it competes on. Notes from the filing.
Anthropic, PBC: Form S-1 registration statement (filed with the SEC on October 26, 2026) Anthropic filed to go public on the Nasdaq Global Select Market under t...
via Hacker NewsUbicloud took Kimi K2.6 tool-call schema accuracy in vLLM from 92% to 100% with constrained decoding. What it does and why vLLM leaves it off by default.
Fixing Tool-Call Accuracy in vLLM (Jesse Donkervliet and Junhao Li, Ubicloud, October 20, 2026) Ubicloud serves open-weight models, and this post explains why a...
via Hacker NewsXGrammar cuts the next-token menu to 2 to 12 tokens on JSON structure. With whitespace off it still expects ': ' after keys unless you pass separators.
Long-running AI agents lock onto one plan. Four October 2026 studies point to one fix: forget and restart. What it costs and what to hand off.
Apollo's chief economist sees no AI payoff outside big tech, and says companies now treat token spend as a cost center. What builders should do.
No Signs of AI Payoff Outside the Tech Sector (Torsten Slok, Apollo's chief economist, The Daily Spark, October 24, 2026) Slok's note is five charts and four sh...
via Hacker News"Our red-teaming agents found vulnerabilities in almost every one of these services that gave them limited forms of internet access."
— OpenAI
Block's Dessa is a full-duplex speech model trained only on human conversation audio. What it gains by skipping text, and what it can't do yet.
Dessa: the Conversation Model trained only on speech (via Hacker News) Dessa is a speech-to-speech model from a small research lab inside Block, announced on Oc...
via Hacker News (295 points on 2026-10-24)"Harness = thing that turns an LLM into an agent Provides tools, context, capabilities etc. Claude Code and Codex are harnesses. Pi is a harness. The thing the LLM is interfaced through is the harness."
— Theo Browne
An unencoded > in the HN Algolia numericFilters param returns HTTP 400 HTML. curl -G with --data-urlencode fixes it.
An agent harness is the software that turns an LLM into an agent: the loop, tools, permissions, sandbox and context management. How it works, with a Python demo.
Harness engineering is the work of building the tools, checks and docs around a coding agent so it can't repeat a mistake. Origins, practices and a lint demo.
Agentic engineering is building production software with coding agents that write and run code. How it differs from vibe coding, with a runnable test loop.
An agent task's time is generation, prefill, tool runs and per-call overhead. A runnable model shows what 1000 tokens per second fixes and what it doesn't.
How to evaluate AI agents: build tasks from real failures, grade outcomes, run several trials and report pass@k and pass^k. With a runnable Python demo.
A system prompt is text placed before the conversation that sets a model's role and rules. See the raw chat template, its token cost and its limits.
An AI agent is a model that calls tools in a loop until a goal is met. A runnable Python loop, what each step costs in tokens, and where agents break.
LLM cost per task is each call's tokens times their prices. For agents, cached input dominates. A runnable calculator with GLM-4.7, GPT-5.2, Sonnet 4.5.
Agent skills are folders with a SKILL.md that agents load in stages. The spec, a runnable demo, and what each stage costs in tokens.
MCP is an open protocol that lets AI apps find and call tools over JSON-RPC. See the four messages a host sends, with a runnable Python server and client.
Claude Opus 4.5 costs $5/$25 per million tokens and adds a beta effort setting. How effort changes token spend and cost per task, with a runnable script.
Reasoning models write hidden thinking tokens before answering, billed as output. How GPT-5.1's reasoning effort works, with a runnable cost script.
Code execution with MCP lets an agent call tools from code, so definitions and big results stay out of context. How it works, with a runnable token count.
Agent sandboxing limits which files an AI agent can touch and which hosts it can reach, enforced by the OS. How Claude Code and Codex do it, with a Python demo.
Cursor 2.0 runs up to eight agents on one prompt using git worktrees. How worktrees work, a runnable best-of-3 script, and the gotchas that bite first.
ChatGPT Atlas puts an agent inside your browser. How prompt injection hides in page text, why filtering it fails, and where the real check has to live.
Durable execution records every step of a workflow so it can resume after a crash without redoing work. How replay works, with a runnable Python demo.
Subagents are helper agents an AI agent spawns with their own context window, tools and prompt. How they save context, what they cost, and how to define one.
AI agent memory is what an agent keeps outside its context window and reads back later. Short-term vs long-term memory, memory tools, and a Python demo.
The lethal trifecta: an AI agent with private data, untrusted content and a way to send data out can be made to leak it. How to spot it and break the chain.
The orchestrator-worker pattern has one LLM break a task into subtasks at runtime, run workers in parallel and merge the results. How it works, with Python.
A multi-agent system splits one task across several LLM agents with separate contexts. When it beats one agent, what it costs, and why handoffs fail.
An AI coding agent is an LLM in a loop with tools to read code, edit files and run tests until a task is done. How the loop works, with a runnable Python demo.
Human-in-the-loop AI pauses an agent so a person can approve, edit or reject an action. Where to put the checkpoints, how to avoid approval fatigue, with code.
The A2A (Agent2Agent) protocol lets AI agents from different vendors discover each other and delegate tasks. How it works and how it differs from MCP.
The evaluator-optimizer pattern pairs a generator LLM with an evaluator that sends feedback in a loop until the output passes. When it works, with Python.
AI agent architectures decide who picks the next step: your code, a plan or the model. Workflows, ReAct, plan-and-execute and multi-agent, with Python.
Deep research agents plan a question, search and read in rounds, then write a cited report. How Gemini, OpenAI and Claude build them, with a Python demo.
Agentic RAG puts retrieval inside an agent loop: the model decides when to search, grades results and rewrites queries. How it works, with a Python demo.
AI guardrails are checks around an LLM call that block, rewrite or route bad inputs, outputs and tool calls. Rail types, limits and a Python pipeline.
Agentic AI means systems that pursue goals over many steps with little supervision. The four dimensions of agenticness and the math of compounding errors.
How to build an AI agent: a tool registry, a model call, a loop with stop conditions, errors and approvals. One runnable stdlib Python file, then the real API.
AI workflows follow code paths you define; agents let the model choose each step. The five workflow patterns, when an agent is worth it, and a cost simulation.
On 22 October 2024 Anthropic put computer use in public beta. Claude scored 14.9% on OSWorld screenshots. Build around the prompt-injection warning.
Tool calling lets an LLM ask your code to run a function with JSON arguments. The request and result shapes, error handling, and a runnable Python demo.
On 13 June 2023, GPT-4 and GPT-3.5 Turbo started returning JSON for functions you describe. The schema, the injection caveat, and an offline check.
Prompt injection is when untrusted text in an LLM's context overrides your instructions. Direct vs indirect attacks, why filters fail, and designs that hold.
ReAct is the agent pattern where an LLM alternates thoughts, tool actions and observations. What the 2022 paper showed, with a runnable Python ReAct loop.