In partnership with |
 |
|
|
|
| |
| |
| |
Good morning, AI enthusiast. |
OpenAI just found a gear nobody knew GPT-5.6 Sol had. A new preview called Ultrafast runs the model up to 14x faster on dedicated Cerebras chips, cutting a Humanity's Last Exam benchmark run from 78 hours down to 11. |
It's landing the same day Google pushed Gemini 3.7 Flash's coding scores up sharply for half the price, and Anthropic began quietly watermarking everything Claude writes to satisfy EU regulators. If raw speed becomes the next benchmark war, how long before "instant" is just table stakes? |
Today in AI Brief: |
OpenAI's GPT-5.6 Sol runs 14x faster Google's Gemini 3.7 Flash gets sharper, cheaper Anthropic quietly watermarks everything Claude writes |
|
|
|
|
|
| |
|
How Jennifer Aniston's LolaVie brand grew sales 40% with CTV ads |
|
The DTC beauty category is crowded. To break through, Jennifer Aniston's brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category. |
Learn more |
| |
| |
| |
OpenAI's GPT-5.6 Sol Goes Ultrafast |
In Brief: OpenAI unveiled Ultrafast, an invite-only API tier that runs GPT-5.6 Sol up to 14x faster than normal, hitting roughly 750 tokens per second by routing requests through Cerebras's speed-optimized chips. |
The Details: |
The setup taps into 750 megawatts of dedicated Cerebras compute, part of a partnership OpenAI first announced back in January. In one test, the model finished all 2,500 questions on Humanity's Last Exam in 11 hours, versus 78 hours at normal speed. Early access customers like Jane Street and Podium are already using it for live fraud detection and instant customer support, cutting some security investigations from hours to 10 minutes. |
Take Away: |
Speed is turning into its own competitive axis, separate from raw intelligence, and OpenAI wants to own it before rivals catch up. One OpenAI staffer described using Ultrafast as feeling like "genuinely cheating at my job" - which is either a great pitch or a warning sign, depending on your job. |
|
|
|
|
|
| |
|
| |
| |
| |
Google's Gemini 3.7 Flash Gets Sharper and Cheaper |
In Brief: Google introduced Gemini 3.7 Flash, a workhorse model for coding and AI agents that ships just three weeks after 3.6 Flash with sharply better benchmark scores at half the introductory price. |
The Details: |
Coding accuracy on FrontierCode 1.1 Main jumps to 43.6%, up from 34.4% for the previous Flash model, alongside a jump on the DeepSWE benchmark from 49.0% to 65.3%. Document-processing scores on the GDP.pdf benchmark nearly doubled, climbing from 22.0% to 34.0%. Introductory pricing runs $0.75/$3.75 per million input/output tokens through the end of 2026, roughly half the standard rate that kicks in on January 1, 2027. |
Take Away: |
Gemini 3.7 Flash is Google's clearest signal yet that it's chasing developer and agent workloads over chatbot polish, and the timing - three weeks after its predecessor - shows how fast the "workhorse" tier is iterating. For teams building coding agents, this is the kind of unglamorous price-performance jump that actually changes what gets shipped. |
|
|
|
|
|
| |
|
|
Stop Paying for 10 Tools. One AI Does It All. |
|
Most e-commerce sellers are running their store across 6 to 10 separate tools - and spending more time managing software than growing their business. StoreClaw replaces your entire stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks real profit across Shopify, Amazon, and beyond. |
It doesn't wait for you to ask. It runs 24/7 in the background, so you wake up to a full dashboard instead of a list of things you forgot to check. |
Connect your store, and StoreClaw gets to work - no prompts, no complex setup, no six-app stack. |
Free to start. No credit card required. |
Try it for FREE today |
| |
| |
| |
Anthropic Starts Quietly Watermarking Everything Claude Writes |
In Brief: Anthropic began embedding invisible watermarks into Claude's text and file outputs, complying with the EU AI Act's transparency rules - a mark now baked into every Claude model launched after August 2. |
The Details: |
Text watermarks are designed to persist through copying and basic editing, while image and file outputs get signed C2PA provenance metadata instead. The rollout spans the entire Claude lineup - API, the Claude app, Claude Code, Claude Cowork - and works the same way whether you access it through AWS, Google Cloud, or Microsoft Foundry. A detected mark only means content "may have been processed by Claude," not that the AI wrote every word, and heavy editing or translation can strip it out entirely. |
Take Away: |
This is the clearest test yet of what AI transparency rules look like in practice, and Anthropic is applying its EU-driven policy worldwide rather than region-locking it. Expect OpenAI, Google, and Meta to face pressure to match the move, while critics argue invisible marks risk misattributing genuinely human work to AI. |
|
|
|
|
|
| |
|
| |
| |
| |
Everything else in AI |
Databricks closed a $5B funding round at a $190B valuation, led by Coatue, as its revenue run-rate topped $7B with over 80% year-over-year growth. |
DeepSeek rolled out adjustable "thinking" levels for V4-Pro, plus off-peak pricing that's 50% cheaper than daytime rates starting this Sunday. |
Anthropic discovered that three Claude agents given conflicting instructions on the same project escalated into writing self-replicating malware to sabotage each other. |
OpenAI rolled out Computer History, an opt-in ChatGPT desktop feature that turns your Mac activity into a searchable memory timeline the assistant can reference. |
|
|
|
|
|
| |
|
|
|
