HackrLife

Kimi K3.1 Pro Closes In on Opus and 86% of Small Businesses Use AI

Plus: can selling to AI agents really bring you users, Shopify's hint-don't-answer loop, 7 ways RLHF goes wrong, a Notion Auto review, 3 startups funded this week, and quick hits.

ChatGPT Ads Reach 89 Countries, and "Build With AI" Hiring Demand Up 51%

Plus: how Anthropic risk-tiers its code review, an IBM paper on why RAG tricks don't stack, Radaric reviewed, 5 startups funded yesterday, and quick hits.

Flint puts a 40B Coding Model on a Laptop, and the median AI paid user pays $25 a month

Plus: why AI won't overtake search in 2026, per a 2025 EMARKETER forecast, cheap-model guardrails that rival frontier judges, Fastlane 2.0 reviewed, and 5 startups funded yesterday.

Otter's $320M Non-Equity Raise and How to Run Your Own AI Credits Programme

Plus: why more memory can make agents worse, Meta's thin-orchestrator pattern, Onepin for AI voiceovers, and 5 startups funded yesterday.

Only 13% of CIOs say AI has paid off

Automated evals that lift accuracy from 78.6% to 90.5% and an open model for text-heavy graphics

Meta Rewrites Its Crawler Terms, the 30-Day Window Where Upsells Work Up to 10x Better, and a Router That Cut Median Agent Cost 64%

Plus: Later's 89% jump in creator payouts, why coding agents fix more bugs with whole files, ngram 2.0 reviewed, and 6 startups funded since Friday.

DeepSeek's $0.75 Search API, 71.5% of AI Tokens Go Open-Weight, and How Reddit Lost Its Grip on ChatGPT's Answers

Plus: the cache trick that cut an agent's cost 4x, why AI judges flip their verdicts, OpenSEO, and 5 startups funded this week.

27M Creator Businesses, the 75% AI Trust Premium, and a new LinkedIn creator tool

Approval gates that make your agent ask before it executes, new research on agents that tune their own harness, and Pancake, the outbound stack your AI agent can run