In partnership with
Welcome to OHMO AI. Google built a Gemini that hunts holes in software, and it isn't handing that one to everyone. GoPro is giving most of itself to a maker of AI data centre parts.
Anthropic cut the price of running agents on its newest models, and the lawsuits blaming OpenAI for a mass shooting have multiplied.
In today's newsletter:
Google ships Gemini 3.8 Flash and a twin that hunts vulnerabilities
GoPro hands 90% of itself to an optics company and heads for AI data centres
Anthropic releases Claude Fable 5.1 and Mythos 5.1 and cuts the cost of running them
OpenAI faces 30 more lawsuits over the Tumbler Ridge mass shooting
LATEST IN AI
Google's new Gemini Flash comes with a twin that hunts vulnerabilities
Image source: Google
OHMO AI: Google announced two versions of Gemini 3.8 Flash: a standard model it calls a "workhorse" for agentic tasks, software development and multi-step reasoning, and Flash Cyber, built to find and mitigate security flaws in code. The standard model is available across Google's products from today, while the cyber version is limited to trusted testers and governments. It's Google's third Flash release in six weeks.
The details:
The model scored 86.2% on CyberGym, a cybersecurity benchmark, and 47.2% on CWE-Bench, which measures how well an AI fixes the flaws it finds rather than only spotting them.
The price holds at $0.75 per million input tokens and $3.75 per million output tokens, the same introductory pricing as Gemini 3.7 Flash, so the jump costs nothing extra.
People can turn the model's effort level up or down for each job, trading quality against cost and how long they wait for an answer.
Why it matters: The price hasn't moved from the last Flash, so the companies behind the apps on your phone can swap in the new model at the same bill, and that's where a regular person feels this first, not in the Gemini app but in other apps getting better at multi-step jobs after a developer flips a switch. Sundar Pichai's own line in an X post was significant leaps over the last Flash in software engineering, agentic tasks and multi-step reasoning. The apps you already use are where that line gets tested.
PRESENTED BY HUBSPOT
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
GOPRO
GoPro hands 90% of itself to an optics company and heads for AI data centres
Image source: GoPro
OHMO AI: The action camera brand GoPro has agreed to merge with Starman Optical, an optical photonics company, in a $285 million deal that hands Starman a 90% stake in the combined business. GoPro's shareholders get about $1.14 a share in cash and keep roughly 10% of a company that now plans to sell Starman's optical transceivers into AI data centres.
The details:
GoPro shares were up 47% on Tuesday morning after the announcement.
Starman's optical transceivers are expected to fall under the combined company's umbrella, which the press release frames as extending GoPro's reach into "the large and rapidly growing market for AI infrastructure in optical transceivers".
The combined company is also planning a push into defence, government, robotics and aerospace, built on its pooled IP, optics and imaging capabilities.
GoPro stays publicly listed, and the cash is paid to shareholders when the deal closes.
Why it matters: The 90% stake says more than the $285 million: the optics maker ends up owning the household name, which reads like a map of where the value in this chain now sits. The AI buildout is recruiting suppliers that were never part of it, and GoPro is the latest unexpected one. If a named customer for those transceivers turns up before the deal closes, the pivot has something under it; until then, the 47% jump reads like the market pricing the buyout.
ANTHROPIC
Anthropic releases Claude Fable 5.1 and Mythos 5.1 and cuts the cost of running them
Image source: OHMO AI
OHMO AI: Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1, two versions of its most advanced model, and paired them with a cut to the price of cached context, the material a model has already read and keeps on hand rather than reading again. The cut is aimed at the cost of running agents, the AI systems that carry out long multi-step jobs on their own.
The details:
Fable 5.1 is the version available to everyone, on cloud platforms and through the Anthropic API from today.
The release includes changes meant to cut false positives from the model's safeguards, the cases where a guardrail blocks something it shouldn't.
Anthropic pitches the cache cut as changing the economics of persistent agents, the ones that keep working over long stretches with the same context loaded.
The new models come with performance upgrades on top of the pricing changes, and the pair is billed as the company's most advanced model to date.
Why it matters: A cut on cached context only pays out for customers who keep the same long context sitting with Anthropic, so it reads like a price for staying put as much as a discount. The deeper your agents run on Claude, the more it costs to move them anywhere else. And the saving is Anthropic's own estimate, made on workloads it defined itself, so the figure that settles it is the one on a customer's next invoice.
PRESENTED BY ATTIO
Introducing The First Agentic CRM
Get revenue agents, workflows, and automations across every stage of your motion. Access customer data in real time through Attio's web app, MCP, API, and SDK.
Then Ask Attio anything about your business and get instant answers.
It's the CRM that runs the work behind every win.
OPENAI
OpenAI faces 30 more lawsuits over the Tumbler Ridge mass shooting
Image source: OHMO AI
OHMO AI: The law firm Edelson PC is filing 30 more complaints against OpenAI this week on behalf of victims and families of the Tumbler Ridge mass shooting in Canada, on top of the seven it filed in April. The new suits accuse OpenAI of aiding and abetting the shooting, rather than just negligently failing to prevent it, and allege that ChatGPT induced the shooter to carry out the attack.
The details:
The new suits were filed in federal court for the northern district of California, on top of the seven Edelson filed in April.
The suits name Sam Altman, OpenAI's chief executive, alongside the company, and accuse both of negligence, aiding and abetting a mass shooting and product liability.
OpenAI's response is that it prioritises safety. The claims are allegations at the filing stage.
Why it matters: Negligence asks whether OpenAI should have stopped a user. Aiding and abetting says the product took part, which is a different kind of claim for a chatbot company to answer. The tell will be whether a judge lets that theory stand: if it does, every company running a chatbot reads as a possible co-defendant in whatever its users do next.
AI GUIDES
Clear a cluttered inbox in ten minutes without deleting anything you wanted
Cleaning an inbox by hand means opening hundreds of promo emails one at a time. With Claude in Chrome open on your Gmail tab, you describe what to flag and what to protect, it sorts the junk by sender with counts, and nothing is deleted until you say yes. You review one list instead of five hundred emails.
Copy this, swap in the bracketed parts, and paste it in.
I am logged into Gmail in this tab. Scan my inbox and flag emails that match: [promotional content, unopened newsletters, shipping notices, SaaS marketing emails]. Leave alone: [personal messages, anything with attachments, emails from my company domain]. Do not delete or send anything yet. Group what you find by sender, with a count for each sender and a one-line note on what they send me. Then show me the full list and wait for my approval. After I approve, I may ask you to delete some senders and unsubscribe from others instead of deleting.Telling Claude what to flag and what to protect keeps the sweep tight, and the grouped list with counts lets you check its judgement before anything happens. Claude always asks before a delete or send goes through, so the only irreversible step is yours.
COMMUNITY
Featured Reader
Every newsletter, we feature how a reader is using AI to work smarter, save time, or make life easier.
Today's reader is Kirsty B. from Hull, United Kingdom:
“Next door's house is the same as ours and their council tax is a band lower. Six years I've paid more and nobody could tell me why.
Photo of the bill into chatgpt on my phone. Asked it what band D means in plain words and whether you can even argue it.
gov.uk band checker: every house on our road, one at a time, pasted into the same chat. Told it: ours and next door's at the top, anything odd in bold.
Same chat: write the challenge to the Valuation Office. Polite, no ranting, same house, here are the bands.
It kept inventing a deadline for appeals that isn't on the gov.uk page, so I cut that line.
Husband says if it goes down he's buying me chips.”
Want to be in the next newsletter? Tell us how you use AI here.
AI Meme
Go Deeper
AI Guides: Practical AI guides, prompts and tools.
Instagram: The AI page they pray you’ll never find.
Facebook: For the ones who never left Facebook.
