This Week in AI — August 11, 2026
Tuesday · This Week in AI · August 11, 2026 · 5 min read
This week was less about one giant headline and more about the ground shifting under the price of doing business with AI. Cheaper access, bigger context windows, and one merger that says a lot about where customer service is headed. Here's what mattered.
1. OpenAI just made AI five times cheaper to run at scale
OpenAI cut the price of GPT-5.6 Luna by 80 percent, down to $0.20 per million input tokens. That's not a small tweak, that's the kind of price drop that changes what's worth automating. A task that used to cost too much to run constantly, like scanning every incoming lead or summarizing every call transcript, suddenly pencils out.
Why it matters: if you talked yourself out of an AI workflow last year because the math didn't work, it might work now. Worth a second look.
2. Google is un-splitting its AI teams
Google is consolidating its AI leadership under one roof, with Demis Hassabis stepping back to chairman and Koray Kavukcuoglu taking over operations. The company has had its AI research and product teams split across two continents since 2023, which has slowed how fast things shipped.
Why it matters: expect Google's AI tools, Gemini included, to move faster and more consistently once one team is steering instead of two.
3. Claude and Gemini are handling bigger jobs without falling apart
Anthropic and Google both pushed task-running agents in front of more users this week, and Anthropic's million-token context window means Claude can now hold an entire year of business documents in a single conversation without losing the thread.
Why it matters: this is the difference between an AI tool that can help with a task and one that can actually sit in on your whole quarter. Feed it more, and it stops forgetting what you told it ten minutes ago.
4. A voice AI company just bought a customer service company
SoundHound AI announced it's acquiring LivePerson, combining SoundHound's voice recognition tools with LivePerson's digital customer messaging platform.
Why it matters: if you've been dealing with separate tools for phone support and chat support, expect that line to keep blurring. One AI system handling a customer whether they call or text is coming faster than people think.
5. AI solved a math problem humans couldn't
A new model reportedly disproved the unit distance conjecture, an open math problem, well enough that a Fields Medalist said he'd recommend the proof for publication in a top journal.
Why it matters: this isn't about math homework. It's a signal that these tools are starting to produce genuinely original work, not just remix what's already out there. Worth watching where that shows up next.
6. The hype is cooling, the useful stuff is heating up
Buyer attention is shifting away from flashy chat demos toward tools that quietly do the boring, repeatable parts of a business: support triage, lead research, document summaries, internal search. Small teams are the ones winning here, not because they chased the newest model, but because they built one workflow and stuck with it.
Why it matters: this is the whole thesis of this newsletter. You don't need the flashiest tool. You need one that handles the thing you dread doing every week.
That last one is the story I'd sit with longest. Everything else moves fast and gets old fast, but "pick one repeatable thing and automate it well" doesn't go out of style.
