This Week in Tech: A Free Frontier Model, Smarter Email, and Big Tech's AI Earnings Test

The third week of July delivered one of the most consequential stretches of AI news this year. A Chinese lab shipped the largest open-weight AI model ever released, one that competes head-to-head with the paid flagships from OpenAI and Anthropic. Superhuman rolled out email auto-drafts that users actually send. Google's next flagship model hit turbulence on its way to launch. ServiceNow's ecosystem quietly expanded AI-powered marketing automation for service organizations. And Wall Street kicked off an earnings season that will test whether Big Tech's massive AI spending is paying off. Every one of these stories has a direct line to how small businesses will run, and compete, over the next twelve months. Let's break them down.

The biggest story of the week came out of Beijing. On July 16, Moonshot AI released Kimi K3, a 2.8-trillion-parameter model that VentureBeat called the largest open-source model ever released, and the first open model to genuinely reach the same tier as the closed flagships from OpenAI and Anthropic. It's live now via Moonshot's API and app, with downloadable weights promised by July 27. Under the hood, K3 uses a mixture-of-experts design: of its 896 expert networks, only 16 activate for any given token, roughly 1.8% of the pool, which keeps running costs far below what the raw parameter count suggests. It also carries a 1-million-token context window, enough to hold an entire client history, contract stack, or CRM export in a single conversation.

What this means for your small business: frontier-level AI is becoming a commodity. When an open model matches the paid leaders, the price of intelligence drops for everyone, API providers cut rates, tool vendors pass savings along, and capabilities that were enterprise-only last year show up in affordable software this year. The businesses that win won't be the ones with access to the best model (everyone will have that). They'll be the ones whose workflows are structured so AI can actually plug in. That's exactly why PointWake starts every engagement with an audit, not a tool recommendation, the bottleneck is almost never the model. It's the process around it.

On July 14, Superhuman launched a revamped auto-draft feature that identifies emails needing replies and writes drafts in your own tone, learned from your past conversations. The numbers from testing are the story: 40% of auto-generated drafts were sent within a day, and 60% of those went out without a single manual edit. That second number matters. AI email drafting has existed for years, but drafts that sound robotic get rewritten, which saves nobody any time. Drafts that sound like you get sent. Superhuman credits the jump to using current frontier models from Anthropic and OpenAI with far more context about your communication style.

What this means for your small business: email is still where most small businesses live, quotes, follow-ups, scheduling, client questions. If your team spends two hours a day in the inbox, tone-matched auto-drafting is one of the highest-ROI automations available right now. But it only works if your email history reflects the voice you actually want. This is a pattern we see constantly at PointWake: the AI is ready before the business is. Clean up your templates and standardize your responses first, and tools like this become force multipliers.

Not every AI story this week was a triumph. Reports surfaced that Google rebuilt Gemini 3.5 Pro from scratch after internal testing exposed structural problems in the original model, targeting a July 17 release that Google itself never officially confirmed. Reported specs, including a context window doubling to 2 million tokens and a new "Deep Think" reasoning mode, remain unverified, and as of this writing the broad launch is still in question.

What this means for your small business: even Google, with effectively unlimited resources, would rather delay and rebuild than ship something that fails in real use. If you're evaluating AI tools for your operation, take the same posture. The vendors promising everything-now are the ones to be most skeptical of. Pilot small, verify results against your own data, and expand what works. A delayed rollout that works beats a fast rollout you quietly abandon in three months.

In quieter but very practical news, marketing-automation platform Tenon announced on July 15 an expansion into the ServiceNow AI Platform, letting organizations run AI-powered, omnichannel marketing on the same data, workflows, and governance that already power their sales, service, and operations functions. The significance isn't the specific product, it's the direction. The industry is collapsing the wall between "marketing tools" and "operations tools." Your CRM, your dispatch board, your invoicing, and your follow-up campaigns are converging into single systems with shared data.

What this means for your small business: if your customer data lives in four disconnected apps, you're becoming the exception, and the market is moving away from you. Unified data is what makes automation compound: a completed job can trigger the invoice, the review request, and the six-month follow-up campaign without anyone touching a keyboard. Consolidating systems is unglamorous work, but it's the foundation everything else sits on. It's the single most common recommendation that comes out of our workflow audits.

Finally, the story hanging over all the others: earnings season opened this week, and as LinkedIn News highlighted Monday, investors are demanding proof that Big Tech's enormous AI infrastructure spending is translating into actual returns. After several quarters of "trust us, it's coming," the market wants receipts.

What this means for your small business: watch this one closely, because it's the same question you should be asking at your own scale. AI spending justified by hype eventually faces its own earnings call. The discipline that protects you is measurement: know what an automated workflow costs, know what it saves in hours or converts in revenue, and cut what doesn't perform. If trillion-dollar companies are being forced to show their math, a small business should be doing the same on day one.

The common thread this week: AI capability is no longer the constraint, implementation is. A frontier model is now effectively free. Email that writes itself is shipping. Marketing and operations platforms are merging. The differentiator left on the table is whether your business processes are organized enough to take advantage.

Here's a practical step for this week: pick the one task your team repeats most often by email, quote follow-ups, appointment confirmations, review requests, and document exactly how it's done today, step by step. That document is the raw material for automation. You can't automate what you haven't defined.

And if you'd rather have a second set of eyes on where AI actually fits in your operation, before you spend a dollar on tools, that's exactly what we do. Free AI Readiness Consult → https://pointwake.com/contact