Tech Digest hero — August 15, 2026

Top story

OpenAI Uncaps Free ChatGPT Text Chats After Cutting Luna's API Price 80%

Source Digital Applied

Why it matters for entrepreneurs: On August 6, OpenAI made GPT-5.6 Luna the default model for Free and Go ChatGPT users, with unlimited text chats and a "Think" button rolling out the week of August 10 — uploads and images stay capped, and the change is subject to abuse guardrails. It's the biggest jump in what a non-paying ChatGPT account can do since the GPT-5.6 family launched, and it lands a week after OpenAI cut Luna's API price 80% (from $1.00/$6.00 to $0.20/$1.20 per million input/output tokens) — almost certainly what made "unlimited" affordable to ship.

For a service business, this changes two things at once. More of your prospects and clients are now doing open-ended research inside a free ChatGPT account instead of hitting a wall after a few messages, so it's worth checking how your business shows up across a longer conversation, not just a first answer. And if your own stack still routes simple client-facing chat through a pricier model out of habit, an 80% Luna price cut is a same-day argument for re-testing that workflow on the cheaper tier before renewing anything.

Quick hits

Gemini Crosses 1 Billion Monthly Users, Becomes Google's Fastest-Growing Product Ever

Source TechCrunch

On August 11, Google said the Gemini app passed 1 billion monthly active users — up from 400 million in May 2025 — making it the 14th Google product to reach that scale and the fastest-growing one yet. It's a distribution signal, not a capability one: two of the three biggest AI consumer surfaces are now both racing to be free and everywhere, worth watching if you're picking a default assistant to build a client workflow around.

DeepSeek Raises API Prices Up to 11x as Demand Outpaces Capacity

Source InfoWorld

Effective August 16, DeepSeek's V4 Flash pricing jumps 57-214% on input tokens and 136-371% on output tokens depending on time of day, with some cache-hit scenarios up more than 1,000%. The company is pushing users toward off-peak hours — 17 of every 24 stay at half price — because it can't keep up with demand. If you or a client leans on DeepSeek for cheap inference, this is the week to check whether the math still works and to build a fallback into your model routing.

Naïve Raises $28.5M to Let AI Agents Run an Entire Business

Source TechCrunch

Naïve packages LLC formation, business email, phone numbers, cloud infrastructure, and connections to Stripe and QuickBooks behind a single API that coding agents can call — it already has 30,000+ developer customers running things like AI automation agencies and, in one case, an autonomous rental-car business. The founder's line worth remembering: for most of these customers, inference cost — not headcount — is now the biggest line item in running an agent-driven business.

Tool / launch watch

Cloudflare launched Kitesurf, a cloud-hosted browser built specifically for AI agents instead of people — no tabs or themes, just fast page rendering, form-filling, and HTML extraction, using less compute than Chromium for common agentic tasks. It's free in beta today through Cloudflare's Browser Run product, so if you or a client is building a browser-automation agent — scraping, form submission, account provisioning — it's worth a same-week test against whatever headless-Chrome setup you're running now.

Source TechCrunch

Angle for the blog

Headline: "Your AI Stack Just Split in Two — Here's How to Audit Both Halves." This week's news shows AI pricing bifurcating into two tracks that service businesses need separate strategies for. Track one is the free-and-cheap distribution tier — OpenAI uncapping free ChatGPT, Gemini crossing a billion users — where the labs are racing to be everywhere and pricing accordingly; that tier is good enough for high-volume, low-stakes client interactions like FAQ bots or intake triage. Track two is the capacity-constrained tier — DeepSeek raising prices up to 11x because demand outstrips supply — a reminder that "cheap AI" isn't a permanent property of any one vendor, it's a temporary condition that changes the moment everyone else notices the same deal. The practical 600-word post: walk an agency owner through auditing their stack in two passes. First, list every place a model does high-volume, low-stakes work — chat widgets, intake bots, first-draft generation — and confirm it's running on the cheapest tier that still clears the quality bar, since OpenAI and Google are actively subsidizing that tier to win share. Second, list every place a model does complex, capacity-sensitive work — agent orchestration, coding, research — and build a fallback model into the routing now, before a provider's next price hike catches you mid-contract. Close with Naïve's inference-cost line: the businesses actually running on agents today already treat model routing as core infrastructure, not an afterthought, and that's the operational shift worth adopting before the next digest's price move forces it.

The Tech Digest is compiled each morning by SyncBroad AI — a plain-English read on AI for service businesses. Browse the full archive, or book a 15-minute demo to see what's actually deployable for your operation.