What’s new in AI, for people using it to get client work done
Updated weekly. Not every AI headline — just the ones that might change which tool you reach for, or what it costs to use it.
OpenAI launches GPT-6.1 Terra, a cheaper, faster model built for everyday work tasks
Why it matters: Terra runs at roughly half the price and three times the speed of OpenAI’s flagship, and it’s pitched specifically for drafting proposals, reading client documents, and editing video — the kind of billable work freelancers do daily. Worth testing on routine client tasks to cut your AI costs.
Read more →ElevenLabs’ launch pricing on its newest voice models has ended
Why it matters: The 72%-off introductory rate on Eleven v4 and v4 Turbo expired, so API pricing jumped back to $0.08 and $0.04 per 1,000 characters. If you quote voiceover, dubbing, or audiobook work using ElevenLabs, rebuild your cost estimates around the higher rate before your next client quote.
Read more →OpenAI, Google, Microsoft, and Anthropic back one shared “do not train” signal for websites
Why it matters: Anyone publishing content — blog posts, courses, lead magnets like your Prompt Pack — can now use a single robots.txt line to tell all four companies not to train on it, instead of juggling separate rules per provider. It’s not enforceable and rollout timelines vary, but it signals that “AI training opt-out” is becoming standard rather than a one-off request.
Read more →Anthropic files to go public on Nasdaq under ticker “ANT”
Why it matters: Anthropic (the maker of Claude) disclosed $16.3B in first-half revenue alongside a large net loss in its IPO paperwork. If your workflow leans on Claude, it’s a sign the company is scaling fast under real investor scrutiny — worth watching for pricing or product shifts as it heads toward a public listing.
Read more →Notion AI’s Agents Get a Major Upgrade — and Get Cheaper to Run
Why it matters: If you use Notion for client workflows, Agents can now run tasks for up to 24 hours with their own browser, and a new “Auto” model router matches Opus 5.5/Astra-level results at roughly 90% lower cost — more automation for less of your tool budget.
Read more →Google Quietly Cuts Free and Plus-Tier Access to Gemini Pro
Why it matters: If you’ve been using a free or $4.99 Google AI Plus account to draft client work or research with Gemini Pro, most prompts now get routed to the lighter Flash-Lite model instead — worth checking your plan before your next deadline.
Read more →ChatGPT’s GPT-6 Brings Faster, More Interactive Answers — Free Tier Included
Why it matters: If you lean on ChatGPT for client research or quick turnarounds, GPT-6 starts answering while it’s still “thinking” (up to 44% faster on search-heavy questions) and can return tappable charts or mini-tools instead of walls of text — and free users get it too, no upgrade required.
Read more →Anthropic Files for Its IPO, Will Trade on Nasdaq as “ANT”
Why it matters: The company behind Claude just took its biggest step yet toward going public — a sign of how much capital is pouring into the AI tools freelancers rely on daily, and a reminder that pricing and roadmaps at these labs can shift fast as they scale toward public-market expectations.
Read more →Anthropic releases Claude Sonnet 5.5 — faster, cheaper, same price
Why it matters: about 30% faster responses and lower typical cost per task than Sonnet 5, at the same price. If anything you use is built on Claude, expect it to feel snappier without you doing anything.
Read more →Google launches Gemini 4 Argon, its new frontier model
Why it matters: Google’s latest flagship model ships with a much larger output limit than before, which matters most for long-form generation — think full articles or scripts in one pass rather than several chained requests.
Read more →OpenAI’s DevDay: GPT-6.1 “Sol,” a new agent called “Dots,” and new pricing tiers
Why it matters: “Dots” is OpenAI’s push into AI agents that act on tasks rather than just answer questions — worth watching if you’re curious whether a general-purpose agent starts eating into what specialist tools (like the ones on this site) do.
Read more →OpenAI delays a separate, more advanced model over security concerns
Why it matters: a reminder that not every AI announcement ships on schedule — some get held back. Worth a grain of salt on roadmap promises from any AI company, including the ones whose tools are on this site.
Read more →