In partnership with
On Friday, OpenAI slowed down Astra because it could not rule out that the model was too good at hacking. On Monday, it shipped a model that is very good at hacking.
In today's issue, HubSpot brings you a database of 100 side hustles for ambitious founders, with startup costs and earning ranges attached.
GPT-5.6-Cyber answers 95% of requests the standard model refuses. Meta dropped a 30B agent model on Hugging Face with no gate.
The lock and the key shipped the same week, from the same industry.
My Coding skills after claude
Special Coverage
In Partnership with Unblocked
[Webinar] Can you prove AI is working?
AI is in your engineering workflow. While the token spend shows it, the throughput doesn't. The human is very much still in the loop, and that's a context problem.
Join live on Aug 19 (FREE) to see:
The 4 metrics to measure the gap where gains leak out before production.
The 8 stages of context maturity, the specific walls capping your metrics, and a free tool to pinpoint where your team is
Why more MCPs and bigger context windows aren’t enough and what it takes to get real value from your agents.
Meta Open-Sources a 30B Agent Model
Meta put Muse Glimmer on Hugging Face Monday under Apache 2.0, a 30-billion-parameter agent model distilled from its closed Muse Spark system. Roughly 4-bit quantization fits it inside a 24GB consumer GPU.
Meta reports 75.5 on MCP Atlas and 51.2 on SWE-Bench Pro, ahead of open models in its size class. It trails them on computer use and terminal work.
Zuckerberg says Muse Spark 1.2 weights follow.
In Partnership with HubSpot
In a world where financial freedom feels like a distant dream, smart women are building wealth on their own terms.
• Finally, a curated database of 100 proven side hustles (that actually work)
• Each idea comes with required startup costs, time investment, and potential earnings
• Exclusive insights from founders who've turned side gigs into 6-figure empires
• Detailed skill requirements so you can match your talents to the right opportunity
• Bonus: Priority scoring system to identify which hustles align with your lifestyle
Don't let another month slip by watching others build their empire. Your next income stream is hiding in our database, waiting to be discovered.
River AI Raises $1.1B Two Months In
River AI, founded by xAI co-founder Igor Babuschkin, raised $1.1B in a seed and Series A led by General Catalyst and AMP PBC. Nvidia, AMD Ventures, Y Combinator and Temasek joined. The company left stealth in June.
River sells post-training as a product: reinforcement learning and LoRA fine-tuning on open models you keep, rather than prompting one you rent.
Lead Analysis
OpenAI Ships a Cyber Model That Rarely Refuses
OpenAI released GPT-5.6-Cyber Monday through Daybreak Red, a vetted tier for authorized vulnerability research and exploit validation. On OpenAI's internal benchmark it completes 95% of advanced cyber requests. Standard GPT-5.6 Sol completes 1.5%.
OpenAI used the model to find two unknown bugs in Chrome's V8 engine, patched by Google as CVE-2026-15903. Accenture, IBM, CrowdStrike, Cisco and Cloudflare have access.
The gap between 95 and 1.5 is a refusal policy, not a capability jump.
Anthropic Stops Renting and Starts Building
Anthropic formed Theseus Infrastructure with Macquarie Asset Management and Singapore's GIC, a platform that will develop, own and lease US data centers with Anthropic as anchor tenant. Macquarie and GIC fund most of the equity per project.
No spend figure was disclosed. Anthropic also agreed to cover consumer electricity price increases caused by the sites.
Quick Hits
ChatGPT lands on Linux. OpenAI put the ChatGPT desktop app into preview on supported Linux systems, covering ChatGPT, ChatGPT Work and Codex. Read more
OpenAI slows Astra. OpenAI paused internal work on its unreleased Astra model after evaluations left it unable to rule out a "Critical" cyber rating under its own preparedness framework. Read more
UK institute reports 19 rogue agent actions. Across 122 cyber evaluation runs between July 25 and 28, the UK AI Security Institute logged 19 unsanctioned actions on the live internet, 17 of them from one model. Read more
A pattern that hides you from cameras. Researchers built an adversarial print pattern that stops surveillance systems from detecting a person wearing it. Read more
