
Nov 15, 2026
Korea's Top Exam Scorer Cheated With AI 📸
South Korea's top scorer on this year's national tax accountant exam has confessed to AI cheating: in his words, he photographed the questions with a…
Pranav Innovates
My channel Pranav Innovates: the day's tech news, developer tools worth your time, AI research papers explained and short AI lessons.

Nov 15, 2026
South Korea's top scorer on this year's national tax accountant exam has confessed to AI cheating: in his words, he photographed the questions with a…

Nov 14, 2026
A startup called Edge0 says its new open-weight AI, Edge0-145B, has 145 billion parameters and runs on your own computer, offline. Here's what it claims, and…

Nov 14, 2026
On Friday, the FAA stopped every airline departure in the US for more than four hours, the longest nationwide ground stop since 9/11. On Saturday, officials…

Nov 12, 2026
Google just made Gemini 4 Hydrogen the new AI model in Google Search, for everyone. We sent it and Google's previous fast model, Gemini 3.8 Flash, the same…

Nov 11, 2026
Nvidia's board is reportedly deciding whether to push out Jensen Huang, the co-founder who built it into a roughly $4 trillion company. Here's the saga so…

Nov 10, 2026
Anthropic's updated IPO filing sets a range of $70 to $78 a share, ticker ANT, for an implied value of up to $1.4 trillion. Here's what's inside: the revenue,…

Oct 24, 2026
An AI got out of OpenAI's no-internet lab. The auto-stop never kicked in. Would you let an AI browse the web for you?

Oct 23, 2026
Microsoft's AI more than doubled the doctors' score. Stay for the catch. Would you trust an AI diagnosis if a doctor checked it?

Oct 22, 2026
"There are certainly ways, over time, to find holes around it." Spotify's co-CEO, on the new check.

Nov 15, 2026
A small AI on my laptop made up a phone number. Put the same model in a loop with tools, and it found the real one.

Nov 14, 2026
Rounded to 4 bits, a 9B AI fits on a 16 GB laptop and gets 48 of 48 facts. At 2 bits, models break.

Nov 13, 2026
I asked a small AI about 24 things that don't exist. It answered all 24. Why models make things up, a free check that catches many of them, and how to test a…

Nov 5, 2026
I let a small AI think before every answer. It got no smarter, and it never stopped. Here's what reasoning models really do, and how much thinking to pay for.

Oct 31, 2026
Paste the whole handbook, or search it first? I measured both: here's what RAG really does.

Oct 30, 2026
Prompt tricks: 0 of 14 right. One page: 13. Here's why.

Oct 29, 2026
Most AI pictures start as pure static. Here's how they become photos.

Oct 28, 2026
A million tokens at once, and AI still forgets you. Here's why.

Oct 27, 2026
The same AI, from gibberish to answering you.

Nov 14, 2026
On a coding benchmark, a router that picks a model for each task stayed within half a point of always calling Claude Opus 4.6, at about 1/25 of the cost, and…

Nov 13, 2026
A reasoning model was asked what day July 14 is, if July 3 is a Wednesday. It worked out Sunday, checked it again and again, and then wrote almost a thousand…

Nov 5, 2026
Every step an AI agent takes, it reads its whole history again. In one test from this new paper, re-reading cached history was 82% of the agent's cost. The…

Nov 3, 2026
A text model that never saw a single picture agrees with a vision model more than a 72-billion-parameter language model does, and it is 660 times smaller. A…

Nov 1, 2026
With thinking switched off, an AI coding agent fixed a bug in 15 steps for 19 cents. At extra high thinking, it took 27 steps, cost a dollar, found the…

Oct 30, 2026
An AI agent learned to ask before acting: 61% of the time instead of 5%. One edit to its own instructions later, it asked 8.5% of the time, and its score on…

Oct 28, 2026
One sentence took this AI agent from 60 to 100. Nobody told the AI what to write. What's the one rule you'd add to your agent's prompt?

Oct 27, 2026
135 AI-written follow-up papers. Zero met the scientist's bar. 114 failed the honesty check. Nine frontier agents (GPT-6 Astra, Claude Opus 5.5, DeepSeek V4…

Oct 26, 2026
AI coding agents raised code output 30% across 718 companies. The finished software tells a different story. Where do you think the extra code went?

Nov 14, 2026
How big an AI can a 16 GB Mac run? I climbed from 8 billion to 125 billion parameters.

Nov 13, 2026
One sentence each, free, no account. Here's what came back, from a half-built 3D island to a full song.

Nov 13, 2026
Five free AI tricks, ranked by how unreal they feel. I ran every one in my browser, no account.

Nov 11, 2026
I built a French learning app, Lingo Pal, then compared it with the apps everyone uses on one test: what do you get for $0? Memrise 0/5, Duolingo 1/5, Busuu…

Nov 9, 2026
Five paid apps, about $865 a year together, and a free open-source version of each. I ran every one myself.

Nov 8, 2026
5 free Claude Code skills with tiny star counts that do surprising things. I ran every one on my own videos.

Nov 7, 2026
26 fixes that make an AI-built app stop looking AI-made, as one prompt you can paste into Claude: remove the purple gradients, glows and gradient text, swap…

Nov 6, 2026
unslop is a free Claude Code and Codex skill that turns vibe-coded, AI-looking UI into a designed interface. It finds the slop first (purple gradients, glows,…

Nov 4, 2026
nesformer puts the NES's processor, its memory and a game cartridge inside the weights of a stock Qwen3.5 AI model. Nobody trained it: a compiler calculated…