the tech roundup*

← Archive

Edition

Sunday · June 21, 2026

6 Top Stories · 3 Under the Radar · 0 Hype · 9 total

Top Stories

TechCrunchCovered by 4 sources

Polymarket reportedly paid creators to fake 1,100+ viral betting wins

A Wall Street Journal investigation says Polymarket paid people to film fake bets and celebrations, identifying over 1,100 deceptive clips, many shot on near-perfect knockoffs of the site (one slip showed the URL "poiymarket.com"). Creators confirmed they were paid but never disclosed it, and none of the winnings were real, with the campaign aimed at US users where its main crypto platform is banned.

So what? ↓

That "normal guy who won big" clip in your feed is an undisclosed ad engineered to make a banned-in-the-US gambling product look like easy money, which is exactly the kind of thing that empties your wallet.

Source: TechCrunch

Google DeepMindCovered by 2 sources

Google's Gemma 4 lands on AWS Bedrock, free to run under Apache 2.0

Google DeepMind's open-weight Gemma 4 family is now on Amazon Bedrock in three instruction-tuned sizes (31B, 26B-A4B, and a tiny E2B), spanning dense and mixture-of-experts setups, all with built-in reasoning, function calling, and text-plus-image input. There's also a separate encoder-free 12B multimodal variant in the lineup.

So what? ↓

Apache 2.0 means you can actually download and run these yourself instead of renting a closed model by the token, which is the thing that keeps your AI bill from being a one-vendor decision.

Source: Google DeepMind

the-decoder.comCovered by 2 sources

Norway is pulling the AI plug for grades 1 through 7

Starting late August, Norway bans generative AI outright for elementary students (grades 1-7) and allows it in secondary schools only under supervision, with PM Store saying kids need to learn to read, write, and do math first.

So what? ↓

While the rest of the world races to stuff a chatbot into every classroom, Norway is betting that an 8-year-old should figure out long division before outsourcing it, and that's a wager plenty of parents will quietly want for their own kids.

Source: the-decoder.com

EngadgetCovered by 2 sources

Ubisoft co-founder Claude Guillemot dies at 69 in a plane crash

Guillemot, who co-founded Ubisoft with his brothers in 1986 and chaired gaming hardware maker Guillemot Corporation, died in a crash that reportedly happened on an airstrip near France's western coast on June 19, per the company.

So what? ↓

He helped build one of the few non-American giants behind games millions of people actually play, from Assassin's Creed to Far Cry.

Source: Engadget

TechCrunchCovered by 2 sources

"In the Weights" is a vanity search, now with an AI score

A new tool called In the Weights bills itself as an AI-centric vanity search and hands you a score, the SEO-rank-checker reflex repointed at AI. Beyond the score itself the material is thin, so what exactly it measures is anyone's guess.

So what? ↓

Vanity metrics are a powerful drug, and slapping a number on "how AI sees you" is engineered to make you refresh it whether or not anyone's told you what it means.

Source: TechCrunch

Under the Radar

◆ Under the Radar1 source · panel-picked

Google teases "Gemini Omni," and that's about all we've got

DeepMind put up a post titled "Introducing Gemini Omni," and the material we have is exactly that: a headline with no benchmarks, no release date, no pricing, and no description of what the thing actually does.

So what? ↓

Until Google attaches some numbers to the name, there's nothing here to judge except the branding, so we're holding this one behind glass.

Source: deepmind.google

◆ Under the Radar1 source · panel-picked

OpenAI files for an IPO, sort of, in the most no-comment way possible

OpenAI confirmed it submitted a confidential draft S-1 to the SEC, the paperwork that precedes a public offering, while explicitly saying it has not set any timing for what happens next. So: the filing exists, the date does not.

So what? ↓

A confidential S-1 is how a company tests the IPO waters without committing, which means the AI lab that powers your chatbots is eyeing public markets but isn't ready to let anyone price the bet yet.

Source: openai.com

◆ Under the Radar1 source · panel-picked

OpenAI publishes a federal AI rulebook, written by OpenAI

OpenAI put out a blueprint for U.S. governance of frontier AI, pitching a federal framework around safety, resilience, and national security. It's the company that builds the models proposing the rules for companies that build the models.

So what? ↓

When the player drafting the rulebook is also the team on the field, it's worth reading every clause twice before anyone calls it a referee.

Source: openai.com