- Everyday AI
- Posts
- Ep 816: ChatGPT Work and GPT-5.6 Sol: What’s New, 5 Overlooked Features and 1 Hot Take
Ep 816: ChatGPT Work and GPT-5.6 Sol: What’s New, 5 Overlooked Features and 1 Hot Take
OpenAI is shutting down Atlas, GPT-5.6 is coming to Microsoft 365 Copilot, and Microsoft is bringing more AI to Windows security. And more.
Everyday AI
July 10, 2026
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
In Partnership With Slack
Meet the All-New Slackbot: Your Personal AI Agent for Work
Most AI tools only know what you type into them. The all-new Slackbot is different, less like a tool, more like a teammate. Built right into Slack, Slackbot already knows your team's conversations, decisions, and the work you've built up over time, so there's no re-explaining yourself.
Slackbot can now read, write, and act across the other apps your team already uses, and it only ever sees what you're already allowed to see, with your data never used to train the model. Check out Slackbot today.
Outsmart The Future
Today in Everyday AI
8 minute read
🎙 Daily Podcast Episode: OpenAI just launched GPT-5.6 and rolled out ChatGPT Work, bringing major upgrades to how AI gets work done. Give today’s show a watch/read/listen.
🕵️♂️ Fresh Finds: Claude gets new in-app browser in Claude Desktop, Anthropic reset Claude usage limits, and NVIDIA unveiled a new AI pipeline for financial data. And more. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: OpenAI is shutting down Atlas, GPT-5.6 is coming to Microsoft 365 Copilot, Microsoft is bringing more AI to Windows security and more. Read on for Byte Sized News.
💪 Leverage AI: OpenAI just launched GPT-5.6 and introduced ChatGPT Work, making AI agents easier for businesses to use. Keep reading for that!
↩️ Don’t miss out: Miss our last newsletter? We covered: OpenAI launched GPT-5.6, Meta is testing AI glasses that continuously see and hear, and SpaceXAI unveiled Grok 4.5. And more. Check it here!
Ep 816: ChatGPT Work and GPT-5.6 Sol: What’s New, 5 Overlooked Features and 1 Hot Take
Of course GPT-5.6 Sol is OpenAI’s best model yet. ☀️
Every new model is.
The real story is what OpenAI did around it.
Codex got a friendlier name, a broader audience, and a much bigger job.
ChatGPT Work is not just another mode.
It is OpenAI merging chat, coding, browsing, files, plugins, and actions into one work super app.
Today’s Everyday AI breaks down what actually changed, what is mostly packaging, and why Anthropic should be paying very close attention.
Also on the pod today:
• GPT-5.6 “Soul” benchmarks drop 🚀
• ChatGPT Work = Codex rebrand?
• Atlas browser quietly disappears 🕵️♂️
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Auriko Reduces inference cost with cache-aware LLM routing, Timbal is the AI platform for enterprise teams, Toyo lives in your messages and handles your inbox, meetings, and follow-ups
Google AlphaEvolve — Google’s AlphaEvolve is now public, helping companies like BASF, Klarna, and JetBrains supercharge code and algorithm optimization with real-world gains.
Anthropic Reset — ClaudeDevs just reset all 5-hour and weekly rate limits for everyone
NVIDIA Finance — NVIDIA’s new pipeline cranked out 500,000 unique financial news headlines by combining global deduplication with evolving prompts.
1X Robot — 1X just showed off NEO’s tendon-driven hands with 25 degrees of freedom, getting closer to human dexterity and strength.
Claude Desktop Updates — Anthropic took a page out of OpenAI’s book with a new in-app browser available Claude Code on Desktop
OpenAI Limits Reset — OpenAI dropped new rate limit resets on GPT-5.6 Sol, but users say it burns through hours and tokens way faster than expected.
Data Center Delays — Record-breaking opposition and delays hit US AI data center projects in early 2026, with billions stalled and grassroots resistance exploding.
SK Hynix Market — SK Hynix is making its big US market debut, aiming to tap into the soaring demand for AI hardware.
AI and Government — The Trump administration is pushing back on state AI laws, arguing that tweaking AI for "ideological ends" could be deceptive.
1. OpenAI is shutting down Atlas and moving its browser tools into ChatGPT and Chrome 🧑💻
OpenAI is closing Atlas, its ChatGPT-powered browser, just months after launch, and shifting its web-browsing agent features into the ChatGPT desktop app and a new Chrome extension.
According to TechCrunch, the move follows an internal push to cut back on “side quests” and suggests OpenAI sees the browser less as a standalone product and more as a set of tools that should live where people already work.
2. OpenAI’s GPT-5.6 is headed to Microsoft 365 Copilot today ⚡️
OpenAI announced today that GPT-5.6 will become the preferred model for Microsoft 365 Copilot across Word, Excel, PowerPoint, Copilot Chat, and Cowork.
The update means Microsoft 365 users should see more capable help with writing, analysis, presentation-building, and team coordination, while Microsoft says the model is designed to deliver better results more efficiently.
3. OpenAI and Google Supplied AI Tools to Blacklisted Chinese Tech Affiliates in Singapore ❌
According to the Financial Times, OpenAI and Google provided advanced AI services to Singapore-based affiliates of Alibaba, Baidu and Tencent, even though their parent companies are on a Pentagon list tied to alleged Chinese military links.
The activity appears to be legal because current U.S. rules do not broadly block Chinese-owned firms from using frontier AI models outside mainland China, but the report is already fueling new calls in Washington for tighter controls.
4. Perplexity adds GPT-5.6 Terra and Sol today ✨
Perplexity says GPT-5.6 Terra and Sol are now live in Perplexity and Perplexity Computer, giving users access to new model options inside its search and agent-style browsing tools.
The announcement quickly drew attention on X, with users focusing less on raw benchmark talk and more on which model can handle multi-step web tasks without losing the plot.
5. Microsoft Pushes AI Deeper Into Windows Security Strategy 🛡️
Microsoft is moving quickly to make AI a bigger part of Windows security, signaling that future protection for PCs will rely more on automated detection, faster response, and tighter system-level defenses.
The shift matters now because cyberattacks are getting more complex, and Microsoft is trying to make Windows feel less like a passive operating system and more like an active security guard.
6. Meta’s AI push sends stock surging 15% this week 🤑
Meta shares jumped Friday, capping a 15% weekly rally, as investors warmed to Mark Zuckerberg’s fast-moving AI strategy after the company released Muse Image and Muse Spark 1.1 in the same week.
The moves signal Meta is trying to catch up with OpenAI, Anthropic and Google while turning AI into more than a costly infrastructure bet, including subscriptions, model access fees and a possible cloud business.
OpenAI just made most enterprise AI stacks look bloated.
GPT-5.6 upgrades the model layer. ChatGPT Work pulls Codex, browser control, local files, connected tools, and scheduling into one place.
Most people will obsess over benchmarks.
The bigger play is OpenAI removing the coding stigma and making agentic work easier for nontechnical teams to use.
That gap matters.
While one company debates chatbot licenses, another can automate SaaS chores, route work cheaply, and supervise agents from a phone.
That’s what we tackled today on Everyday AI: why the rename matters, where the super app unlocks business value, and why Terra may threaten Anthropic more than Sol.
1. Make Codex feel like work ⚡
ChatGPT Work on the web is mostly a cleaner front door to abilities already hiding inside ChatGPT.
On desktop, it’s Codex with less technical clutter and a name that won’t scare off non-coders.
That sounds cosmetic. It isn’t.
Power users missed scheduled tasks because they were buried, while everyone else assumed Codex was built for developers.
OpenAI fixed the perception bottleneck without rebuilding the engine.
Your rollout can stall the same way. If the interface or language makes capable employees feel unqualified, adoption dies before useful work begins.
Try This: Put your most capable AI workflow in front of five nontechnical employees without coaching them first.
Ask where they’d start and which words feel off-limits. Then rename the workflow around the business outcome, not the technology underneath it.
2. Automate the 12-click browser work 🚀
The desktop merge is where the packaging starts earning its keep.
ChatGPT Work can use local files, span desktop apps, open multiple browser tabs, retain logins, and show computer use in picture-in-picture.
Remote lets users steer tasks from mobile, while a Chrome side panel keeps long-running work visible without constant switching.
That attacks the work most automation pitches ignore.
Real enterprise processes crawl through old websites, password screens, downloads, uploads, spreadsheets, and 12 mind-numbing clicks somebody repeats every morning.
Connectors won’t cover all of it.
Browser control reaches the ugly gaps where APIs don’t exist, while ChatGPT Sites can turn dashboards, trackers, and reports into one live source of truth.
Now the super app idea makes sense.
Try This: Pick one recurring workflow that crosses at least three logged-in tools and burns 30 minutes or more.
Document every click, wait, handoff, and approval. Then have the desktop agent reproduce it with human review before any irreversible step.
3. Benchmark Terra before paying premium 🔥
The flashy comparison is Sol versus Claude Fable 5.
The expensive mistake is ignoring Terra.
In the Artificial Analysis comparisons we covered, Sol was essentially tied with Fable 5 on intelligence while leading on the coding and agentic indexes.
Then the medium model got rude.
Terra ranked second on the coding index, ahead of Fable 5, while costing less than one-fifth as much for the same task.
That pressures premium-model spending because a flagship win matters less when a balanced model clears the work for far less.
Enterprises don’t need the smartest model on every prompt.
Choose the least expensive option that clears the quality bar without creating enough cleanup to erase the savings. Otherwise, the company is buying luxury intelligence to do office chores.
Try This: Build a 20-task evaluation from work your team already pays people to complete.
Include one spreadsheet, one browser process, one research brief, one presentation, and one long-running project. Run Sol, Terra, and your current default against the same rubric, then compare quality, cleanup time, and accepted-output cost.
Route by task difficulty, not vendor loyalty.
Reply