Future Technology[AI Weekly] So, OpenAI shipped three models this week
So, OpenAI shipped three models in one week. Not an update, three separate products with three separate names, because apparently one flagship isn't a big enough decision for anyone any more.
The Big 3
GPT-5.6 splits into three: pick your price, pick your speed
OpenAI's GPT-5.6 family went public on 9 July after a customer-by-customer government review, and it's now ChatGPT's default. Sol is the flagship, built for agentic coding, science and cybersecurity work with a new Ultra subagent mode. Terra targets GPT-5.5-level quality at half the cost. Luna is the fast, cheap tier for high-volume tasks. It's OpenAI finally admitting what anyone building on its API already knew: most tasks don't need the most expensive model, they need the cheapest one that's good enough.
Why this matters: if you're paying for GPT-5.6 API access, Terra probably covers 80% of what you were using the flagship for at half the price. Worth auditing your usage this week.
(3 minute read) Source · Read more on Future Technology →
Illinois just wrote the toughest AI safety law in America
Governor JB Pritzker signed the Artificial Intelligence Safety Measures Act on 6 July, requiring model developers to publish a framework explaining how they assess "catastrophic risk," defined as anything that could kill or seriously injure more than 50 people or cause over $1 million in damage. It follows California and New York down the same path, and it lands while the FTC is separately taking public comment (through 31 July) on whether states can even force AI models to alter their "truthful" outputs. Washington and the states are pulling in different directions at exactly the same time.
Why this matters: if your business relies on a US-based model provider, expect more compliance paperwork and possibly region-locked features as this state-by-state patchwork keeps growing. As of July, states have enacted 109 AI-specific laws.
(4 minute read) Source · Read more on Future Technology →
DeepSeek wants to stop asking Nvidia for chips
DeepSeek is reportedly building its own AI chip, holding talks with chip-design, foundry and memory partners after roughly a year of quiet hiring. It's not alone: Zhipu is doing the same, and the news gave the whole Chinese AI-chip sector a lift as Huawei's grip on the domestic market loosens. This is the bit nobody predicted after the original DeepSeek moment eighteen months ago: the efficiency story is turning into a hardware independence story.
Why this matters: a chip-independent Chinese AI stack changes the leverage US export controls actually have. Worth watching if your supply chain touches AI hardware anywhere.
(3 minute read) Source · Read more on Future Technology →
Quick Hits
Grok 4.5 undercuts Opus on price. xAI shipped Grok 4.5 on 8 July, pitching it as "Opus-class, but faster and cheaper." Whether it holds up under real workloads is still an open question, but the pricing pressure on Anthropic and OpenAI is real. (2 min read) Source · Read more →
Five governments agree AI agents are a security headache. The US, UK, Australia, Canada and New Zealand's cyber agencies jointly published guidance on agentic AI risks this month. Good to see someone thinking about this before an agent with banking access gets talked into a wire transfer. (2 min read) Source · Read more →
Alibaba's cloud business is quietly an AI company now. AI products made up 30% of Alibaba's external cloud revenue this quarter, its eleventh straight quarter of triple-digit AI growth. While everyone watches the US labs, this is where a huge chunk of enterprise AI spend outside the West is actually going. (2 min read) Source · Read more →
Anthropic keeps extending free Fable 5 access. Now pushed to 19 July, the third extension in five weeks. Read into that whatever you like about the pressure Anthropic's feeling with Gemini 3.5 Pro's launch landing the same week. (1 min read) Source · Read more →
Tool of the Week
Canva Grow 2.0
Bundles AI ad creation, publishing, reporting and creative refresh into one Canva workflow, across Meta, TikTok and LinkedIn.
If you're running paid social for a small brand or a client without a dedicated ads team, this collapses four separate tools into one tab.
Caveat: it's still Canva-quality templated output, don't expect agency-grade creative out of it. No affiliate programme set up with Canva yet, so this is a straight recommendation, not a paid link.
One More Thing
New AI models are now arriving somewhere in the world roughly every three days. At that pace, "did you see the new model?" stops being a conversation starter and becomes background noise. The actual skill this year isn't keeping up with every release, it's knowing which three actually matter to you.
Forward this to someone who'd find it useful. Or don't. But you should.
Free weekly AI briefing: futuretechnologyhq.com/newsletter
Some links in this newsletter may be affiliate links. We only recommend products we genuinely think are worth your time.