Do More Newsletter

This issue contains featured article "How Good Are AI Translations, Really?" and exciting product information about Periscope from Duet Display, Zencoder Zenflow Work, Shutterstock app in ChatGPT, Kikoff Fynn, and Visa Intelligent Commerce Connect.

Sponsored by

Keep up to date on the latest products, workflows, apps and models so that you can excel at your work. Curated by Duet.

Stay ahead with the most recent breakthroughs—here’s what’s new and making waves in AI-powered productivity:

Periscope, the innovative new VPN from the makers of Duet Display, is redefining what secure remote access can look like. By connecting your device directly to your home computer and combining that with powerful end-to-end encryption, it offers a level of privacy and control that traditional VPNs simply can’t match. It lets you appear as though you’re browsing from home, unlocking seamless access to your usual apps, services, and workflows no matter where you are. If you care about security, flexibility, and a smoother remote experience, this is one product you’ll want to learn more about.

Zencoder launched Zenflow Work, a desktop orchestration platform built to help people coordinate the work around coding, not just the code itself. It is designed to streamline planning, reporting, communication, and business app workflows across product, marketing, sales, customer success, finance, and HR, which makes it relevant for both small teams and solo operators who want less manual busywork.

Shutterstock launched its app inside ChatGPT, letting users find images, video, music, and sound effects from its licensed library directly in an AI workflow. That matters for small businesses, marketers, and creators who need commercially safe content fast, because it reduces the time spent bouncing between tools and makes asset discovery more conversational and efficient.

Kikoff launched Fynn, an AI credit coach built to help consumers understand their credit journey and get personalized guidance based on their financial profile. It stands out because it translates AI into practical help for budgeting, credit improvement, and next-step decision-making rather than abstract chat.

Visa unveiled Intelligent Commerce Connect to help businesses participate in AI-driven shopping and agentic commerce more securely. For small businesses, the practical upside is discoverability and checkout support inside AI platforms, which is important as more consumers start using AI assistants to shop, compare, and buy.

As remote work and digital privacy take center stage, the makers of Duet Display have introduced Periscope which is an exciting new VPN product. Designed for both everyday users and professionals, Periscope offers a fresh approach to staying connected securely from anywhere in the world. Rather than relying on traditional VPN infrastructure, it brings a more personal, high-performance solution that puts control back in the hands of the user.

What makes Periscope stand out is its unique direct connection model. Instead of routing your internet traffic through distant servers owned by a VPN provider, Periscope connects your remote device directly to a computer on your home network. This means your data travels along a private, dedicated path that you control, significantly reducing the risks typically associated with third-party routing and shared infrastructure.

Security is at the heart of Periscope’s design. By combining this direct connection with state-of-the-art end-to-end encryption, the platform ensures that your data is completely shielded from prying eyes. There is no opportunity for intermediaries to intercept or monitor your activity—your connection is entirely private from end to end. For users concerned about surveillance, data leaks, or unsecured public networks, this level of protection is a game changer.

Another compelling advantage of Periscope is its ability to make your online presence appear as though you are browsing from your home network, no matter where you are in the world. This opens the door to a range of benefits, from accessing region-specific content and services to maintaining consistent access to work systems that rely on trusted network locations. It eliminates the friction that often comes with traveling or working across different regions.

With Periscope, Duet Display’s team has re-imagined what a VPN can be. Faster, more secure, and more intuitive. By eliminating unnecessary middlemen and focusing on direct, encrypted connections, Periscope delivers a modern solution for privacy-conscious users. Whether you’re protecting sensitive business data or simply want a safer way to browse on the go, Periscope offers a powerful new standard in secure connectivity.

How Good Are AI Translations, Really?

A few months ago, I got into a Lyft and realized my driver and I didn’t share a language. He spoke Spanish, and I don’t know enough Spanish to say I don’t know Spanish. Still, I had some specific instructions on where I wanted to be dropped off.

So I did what people do now: I pulled out my phone, opened ChatGPT, and typed what I wanted to say and asked it to translate into Spanish. Then I just held up the phone to my driver. He read it and dropped me off where I wanted.

It was, honestly, kind of magical. Like we’re at the early stages of a universal translator like from Star Trek.

But here’s the thing that stuck with me afterward: There is quite a bit of trust here. That I was dropped off where I asked seemed to suggest the translation was right, but I couldn’t really be sure.

And that question — how much should I actually trust this? — turns out to be the question a lot of people are asking right now.

Because AI translation has gotten shockingly good. But “shockingly good” and “reliable” are not the same thing.

The Numbers Are Impressive (With a Catch)

Let’s start with what the data says, because the headline numbers are genuinely striking.

It’s hard to slap a single “accuracy percentage” on AI translation — it depends too much on the language, the subject matter, and how you measure. But two things are clearly true: mainstream tools now cover 100+ languages (Google Translate claims 249), and in head-to-head evaluations, the big AI models are routinely beating traditional translation engines.

In the WMT 2024 General Machine Translation shared task — the main annual competition where translation systems get tested against each other — Claude 3.5 Sonnet was the best overall system, winning in 9 out of 11 language pairs. A follow-on study called WMT24++ expanded that testing to 55 languages and dialects, and frontier LLMs ranked highest for every single one — though the researchers caution those results are based on automatic metrics and still need confirmation through large-scale human evaluation.

Separately, the localization platform Lokalise ran a 2024 blind-comparison study with 600+ pairwise human evaluations. LLM translations were rated “good” in 56–80% of cases depending on the language pair, and Claude 3.5 Sonnet was preferred in 78% of head-to-head matchups.

These are not small numbers. Five years ago, machine translation was the thing you used when you wanted to get the gist of a menu in another country and didn’t mind if “chicken breast” came out as something anatomically creative. Today, the output reads like a competent human wrote it.

But here’s the catch those numbers obscure: translation errors are not evenly distributed. AI translation tends to fail in exactly the places where failure matters most.

Where It Breaks

Talk to anyone who works in professional translation and they’ll tell you the same thing: the problem isn’t that AI translations are bad. The problem is that they’re confidently wrong in ways that are hard to catch.

A few patterns show up again and again.

Idioms and figurative language. If you tell someone in English, “We really dropped the ball on this one,” AI might translate that into Spanish as a literal statement about dropping a ball. Research from Appen on culturally nuanced translation found that LLMs routinely mistranslate idioms and puns across languages, often requiring human intervention to catch the errors. The sentence comes out sounding fine, just... missing the point.

Cultural context. Language isn’t just vocabulary and grammar. It carries culture. Formality levels, humor, religious connotations — these shift meaning in ways AI doesn’t always catch. Amazon learned this the hard way when it launched in Sweden and “rapeseed oil” got translated into something... significantly more offensive (the Swedish word for “rape” as in the plant shares a root with the word for sexual assault). That was a traditional MT system, but LLMs still stumble on the same class of problem. They produce output that’s linguistically correct but culturally tone-deaf.

Tone and nuance. Sarcasm, hesitation, softening language, implied meaning — these are hard even for bilingual humans. For AI, they’re essentially invisible. The translation might be technically accurate and completely miss the emotional register of what was actually said.

High-stakes terminology. Medical dosages, legal clauses, contract terms, safety warnings. When a mistranslation of “take twice daily” becomes “take two at once,” the consequences aren’t academic. A 2025 study in JAMA Network Open found that AI-translated hospital discharge instructions performed reasonably well for Spanish, but were significantly worse — with far more clinically impactful errors — in Simplified Chinese, Vietnamese, and Somali. The takeaway: performance varies sharply by language and document type, and in medical contexts, “pretty good on average” can still mean dangerous for specific patients.

The Church Question

My own church has been exploring AI translation for live services — specifically, translating our English sermons into Spanish in real time for members of our congregation who are more comfortable in Spanish.

It turns out we’re not alone. There’s a whole ecosystem of tools built specifically for this: platforms like Wordly, Glossa, and Church Translation Live that offer real-time AI translation in dozens to 100+ languages, depending on the platform. Congregants scan a QR code with their phone and get live captions or audio in their language as the pastor speaks. No interpreter booth, no delay, no scheduling headaches.

On paper, it’s incredible. It solves a real problem — one that many churches have struggled with for years using expensive human interpreters or just... not solving it at all, leaving Spanish-speaking members to sit through a service they can only partially follow.

But sermons aren’t product manuals. A pastor might say “God doesn’t just want your Sunday best” — meaning God wants authenticity, not performance. Will the AI understand that “Sunday best” is a metaphor for putting on a show, not a reference to clothing? Maybe. Will it convey the warmth and the gentle challenge in the pastor’s tone? Almost certainly not.

There’s also the question of theological precision. Religious language is loaded. A word like “grace” carries centuries of meaning that shifts between traditions. “Spirit” can mean a dozen different things depending on context. When you’re translating something that matters deeply to people — something they’re building their lives around — “pretty good” isn’t always good enough.

So Where Does That Leave Us?

Here’s how I’d frame it after spending time with the research and, more importantly, after using these tools in the real world.

For casual, everyday communication — AI translation is genuinely great. My Lyft ride worked. Ordering food abroad works. Texting with a colleague who speaks another language works. The stakes are low, the context is clear, and “close enough” is more than enough. If the alternative is no communication at all, AI translation is transformative.

For anything high-stakes — trust but verify. Medical information, legal documents, important business communications, religious services. AI can give you a strong first draft, but it needs a human eye. One practical trick: running translations through multiple AI models and comparing the output. When systems disagree, that’s where the errors hide. Early internal evaluations of this consensus approach have reported roughly 18–22% fewer visible errors in business and legal material — not a magic bullet, but a meaningful improvement if you’re building a workflow around this.

For cultural and emotional nuance — AI still isn’t there. It might get the words right and miss the meaning. This is the gap that will take the longest to close, because it requires something closer to understanding than pattern matching.

The honest summary: AI translation in 2026 is the best it’s ever been by a wide margin, and it’s still not good enough to blindly trust when the stakes are real. That’s not a contradiction. It’s just where we are.

And I think the real skill — the one worth developing — isn’t learning to use the tools. It’s learning when to trust them, when to double-check them, and when to pick up the phone and call someone who actually speaks the language.

Partner Spotlight: Duet Display

Duet Display continues to push the boundaries of productivity and connectivity with its innovative software solutions designed for modern work environments. Known for transforming iPads, tablets, and additional PCs or Macs into high-performance secondary displays, Duet Display empowers users to work smarter across platforms and locations. With a strong focus on performance, low latency, and security, Duet’s technology is trusted by professionals, creatives, and businesses alike. As remote and hybrid work evolve, Duet Display remains at the forefront—delivering seamless, reliable tools that enhance flexibility while maintaining a premium user experience. Find out more at Duet Display.

What Will Your Retirement Look Like?

Retirement looks different for everyone. What it costs, where the income comes from, how long it needs to last. Those answers are specific to you.

The Definitive Guide to Retirement Income helps investors with $1,000,000 or more work through the questions that matter and build a plan around the answers.

Download your free guide to start turning a savings number into an actual retirement income strategy.

Stay productive, stay curious—see you next week with more AI breakthroughs!