AI News, page 21 of 81

AI and software engineering, as it happens.

OpenAI's Aura safety check can stop ChatGPT and Codex from opening a siteSearch Engine Journal

Ahrefs tracked 37,500 AI citations for 20 days and half vanished each dayAhrefs

Google Research proposes a sleep phase so LLMs keep learningGoogle Research

LangChain released a Slack design agent that costs 16 times lessLangChain

70 percent of 100 million dollar Series A rounds went to AICrunchbase News

HubSpot lost 80 percent of its blog traffic and revenue still grewSearch Engine Journal

Bot Food raises $1.5M for a personal context library for AIBetaKit

Locals runs many coding agents on a Mac, three times fasterGenerated Computers

Blake Anderson posts his stack for profitable AI apps, 569 starsBlake Anderson

MicroZig bans every LLM use, including spelling fixes and bug findingMicroZig

Grim Prompter turns a coding agent into a threat modelerGitHub

GitHub releases lever, a Copilot plugin that writes feedback loops firstGitHub

Lantern AI runs recruiting agents that sourced for Crypto.com, Hyatt and Baker McKenzieY Combinator

GrayPass checks who is in control before an agent moves money or exports dataY Combinator

TrueMile is a trucking company run by AI, with 17 percent higher revenue per mileY Combinator

AgentFlux is a self-hosted visual workflow builder for AI agentsGitHub

Kanzo is a kanban board for you and your coding agents, in one 4 MB binaryGitHub

DeepTutor v1.7.0 is a self-hostable AI tutor with 48,000 starsGitHub

MiniMax open-sourced the skills its own coding team uses day to dayGitHub

Hex is a hold-a-shortcut voice dictation tool for macOS, open sourceGitHub

Taste Skill at 104,000 stars gives coding agents a design checkGitHub

Claude Code quietly spends 16 percent of your tokens on a status lineFlamerix AI

Google says AI Mode had a bug that ducked 2020 election questionsArs Technica

CaseCite is a self-hosted open-source take on Harvey, with 69 HN pointsGitHub

Alphabet says AI Mode made less money per search than classic resultsSearch Engine Journal

Replit's ViBench 2.0 benchmark asks AI agents to build whole productsReplit

Upwind raised $400M at $3.8B after walking away from DatadogTechCrunch

Replit says its microVM sandbox blocked the Linux kernel exploitsReplit

Microsoft plugin cuts coding agent cost 18 to 31 percent on SWE-BenchMicrosoft Research and UC Berkeley

Agents start 40 percent of GitHub pull requests, up from 17 percentGitHub

Archive

2415 items published so far, by month.

November 2026463 October 20261134 September 2026765 August 202653

Start a project