|
|
|
Well, ChatGPT had a big day yesterday! The biggest undersold news of the day was the fact that Codex got rebranded to ChatGPT Desktop (I think we were the first ones to confirm this btw??). | After the smoke cleared, we saw some posts about how that might be a generational branding fail, as Codex was really on a brand-equity building streak there for a few monthsā¦ but I mean, cāmon, how are you gonna beat ChatGPT with roughly 1B weekly active users on brand? As dumb of a name as it isā¦ | Not only that, ppl are confused. Both Ethan Mollick and Simon Willison literally said theyāre confuzzled by the differences in the ChatGPT for Work vs Codex, and these are two guys who follow this space pretty darn closely. Uhhh OpenAI marketing?? Plz help? | What do you think? Should they have rebranded ChatGPT to Codex, or vice versa? The latter wouldāve been a major upset, but I think that would be my preference. HBU? | Should they have called the SuperApp ChatGPT Desktop or Codex?Which do you prefer? Pick an option below and use the "Additional Feedback" box to weigh in. | | More on what changed below. | Hereās what happened in AI today: | šŗ OpenAI released GPT-5.6 and ChatGPT Work, turning ChatGPT into a desktop work agent for planning, browsing, files, and deliverables. š° Meta released Muse Spark 1.1 with a 1M-token context window, computer use, coding gains, and lower pricing than many frontier rivals. š° China weighed a āsilicon curtainā around advanced AI models as U.S. companies increasingly adopted cheaper Chinese systems. š° More than $130B in U.S. AI data-center projects were blocked or delayed over local power and water pushback. š” AI 2040: Plan A sketched a hopeful AI future built on transparency, competition, coordination, robot fees, and a citizenās dividend.
| ā¦and a whole lot more that you can read about here. | Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us!Ā | | š¼ OpenAI Launched GPT-5.6 and Turned ChatGPT Into a Desktop Super-App | So, if youāre a ChatGPT user, thereās a high likelihood you just opened the desktop app yesterday or this morning and found a whole new model family waiting for you, along with a totally different vision of what ChatGPT is supposed to be. | See, OpenAI just officially launched GPT-5.6, and then surprisingly also rebuilt the app around āChatGPT for workā to enable āworkā that continues across chats, files, browsers, and devices. And like we said, itās surprising and upsetting a many engineers. | Here's what happened: | To review, GPT-5.6 comes in Sol, Terra, and Luna: think of them like most powerful, balanced, and affordable. Sol's Ultra mode coordinates multiple agents on harder jobs. The new ChatGPT Work can browse the web, use connected apps, edit files, operate your computer, schedule tasks, and create slides, spreadsheets, documents, and sites easier and better than ever before. Codex is moving inside the unified desktop app, while OpenAI sunsets its early browser experiment Atlas and shifts those browser features into the app and a Chrome sidebar.
| People's reaction: | Tibo Sottiaux called Sol OpenAI's best model yet across coding, difficult context, strategy, and site building. Andrew Curran highlighted OpenAI's claim that Sol helped post-train Luna. Simon Willison praised the capability while documenting a dizzying matrix of three models, multiple effort settings, prices, tool calling, and multi-agent options. Jun Song called Sol strong but incremental. Peter Gostev saw Fable as the āwise owlā for architecture and Sol as the dependable executor. And yet: Theo, signĆ¼ll, Corbin Braun, and Daniel Lockyer all captured the UX problem: early Codex users liked their more a focused Codex tool and instead received a super-app with more modes & UI changes to understand. And yet Sriram Krishnan argued that this super-app is where the market is heading.
| How to try it: | Update or download the ChatGPT desktop app. Pick ChatGPT for Work or ChatGPT Codex based on the job (normal vs coding). Choose Sol, Terra, or Luna, then raise effort when the task deserves it. To chat, select āChatā and you can chat anywhere with the floating chat window.
| Just donāt ask us where your chat threads are nowā¦ that part is confusing as all get out. | Why this matters: Grok released a new coding model. Meta just released a new coding model. Gemini 3.5 is supposedly coming out July 17. Presumably, all of them will be cheaper than Fable (Sol is supposedly 1/3 the cost). This is Anthropicās game to lose. | Our take:Ā Letās talk about UX, bay-beee. Once again, the UI problem is really a memory problem. Everyone is confused about where their chats went in the super-app, which is fair. But surfing through old threads are miserable to reopen and skim. IF I never had to scroll anything ever again it would be too soon. Can voice / AI UI totally remove scrolling from my life plz? Itās the worst UX ever, IMO. Your information should come TO you now. | Besides killing scrolling, what would be better UX is really just one chat. One chat that remembers everything important about your projects, tools, and decisions, thatās kinda always running and easy to pull up, that can surface info as needed, and sure, maybe it floats around wherever you go. | That appears to be OpenAI's goal. It is not a bad pattern. Itās just different. But until memory becomes reliable enough to replace a thousand chat threads with one thread to rule them all, the app will continue to feel like dude whereās my chat because somebody rustled all the papers around on your desk and now you canāt find anything. | Hey OpenAI, my partnerās a professional organizer, in case yāall need some tips on Marie Kondoing your SuperApp! | |
|
|
| Claude is currently the most powerful tool of 2026. It's been launching new features every week- Skills, Connectors, Cowork, vibe coding. Yet almost no one knows how to actually use them. | Our expert mentors have condensed 800+ hours of Claude research, articles, YouTube content and real-world practice into a focused 16-hour curriculum. Join the 2-Day Claude AI Mastery Workshop: a live, end-to-end deep dive into Claude plus 10+ AI tools, LLMs and workflows. | It would be silly not to SIGN UP | The professionals who master Claude right now will design carousels in seconds, ship code without being developers, and have AI working for them while they sleep. | You will learn how to: | - master Claude's three modes : Chat, Cowork and Code. - Set up Skills, Connectors and Plug-ins to automate your desktop, Notion and files. - Vibe code apps and dashboards without writing code & 10+ AI tools and workflows that pair with Claude.
| š§ Saturday & Sunday | š 10 AM ā 7 PM EST | Register NOW! | |
š AI Skill of the Day: Use One Smart Model With Cheaper Workers | Using your smartest model for every token is like hiring a CEO to update every spreadsheet cell. Anthropic shared two patterns for keeping Fable 5 in charge while Sonnet 5 handles most of the expensive, token-heavy work. | Use Fable as an advisor. Sonnet executes the task, then calls Fable only when it needs strategic guidance or a course correction. Anthropicās advisor tool gives the advisor the full conversation, then returns its guidance to the executor. On SWE-bench Pro (a recently criticized coding benchmark), this pairing reached about 92% of Fableās score at roughly 63% of the price. Use Fable as an orchestrator. Fable makes the plan and delegates execution to Sonnet sub-agents. Anthropicās cookbook shows the āplan big, execute smallā setup. On BrowseComp (a browser benchmark), it reached 96% of Fableās performance at 46% of the price.
| So use the advisor pattern for one hard task that occasionally needs steering. Use the orchestrator pattern when the work can be split across multiple agents. Each sub-agent keeps its own cache, so repeated calls do not repay the full context cost. | Act as the lead planner. Break this task into clear work packages. Delegate routine, token-heavy execution to lower-cost workers. Keep strategic decisions, quality checks, and course corrections for yourself. Review each workerās result before producing the final answer.
Task: [PASTE TASK]
| Weāll get you some more GPT 5.6 tips on Sunday after weāve had time to digest it all! | Have a specific skill you want to learn?Ā Request it here.Ā | | |
| ChatGPT Work gives you a GPT-5.6-powered work agent that can plan, browse, use connectors, generate deliverables, schedule work, and continue across desktop and mobile; no pricing details beyond plan availability. Meta Muse Spark 1.1 gives developers a 1M-token model for coding, tool use, computer use, and multimodal reasoning through Metaās public API; pricing starts at $0.80 per million input tokens and $3.20 per million output tokens. Claude Reflect shows private Claude usage reports and visualizations across 1/3/6/12-month windows, maps habits to the 4D AI Fluency Framework, and adds quiet-hours nudges; beta, no pricing details beyond Claude availability. Reve 2.1 generates and edits images with stronger prompt adherence and design control, giving creative teams another high-quality visual model to test; no pricing details. OpenKnowledge gives you a local-first markdown workspace where people and agents can co-edit knowledge bases with real-time collaboration, MCP connections, git-backed sharing, agent skills, and search over your notes; no pricing details.
| |
|
 | Initial thoughts and impressions on the news and earliest models |
| To watch the recap of todayās livestream, click the image above or read the catch-up quick version here. | | š° Around the Horn | | Meta released Muse Spark 1.1 through the new public Meta Model API, with 1M-token context, tool and computer use, coding gains, multimodal reasoning, and pricing pressure on OpenAI and Anthropic. More than $130B in U.S. AI data-center projects were blocked or delayed in a single quarter because of local pushback over power and water use. OpenAI published national-security principles allowing defensive cyber, biosecurity, and allied government work while rejecting mass surveillance, autonomous weapons direction, and high-stakes automated decisions without human judgment. AI super PACs raised more than $200M to shape national AI regulation, while Marc Andreessen joined a Federal Reserve AI task force examining productivity, jobs, and the economy. Patreon partnered with Cloudflare to block AI training crawlers from creatorsā work across the platform, with CEO Jack Conte calling for credit, compensation, and consent.
| |
|
|
| Want a computer you can actually train your own AI on? Dell Pro Max with GB10 brings NVIDIA Grace Blackwell architecture, 128GB memory, and NVIDIA DGX OS 7 to founders, builders, and AI enthusiasts testing agents, demos, and model workflows locally. | Check out the Dell Pro Max with GB10 here. | |
š” Intelligent Insights | | AI 2040: Plan A: the ppl who wrote AI 2027 released their recommendation for an optimistic AI future via policy: more transparency, multi-lab competition, international coordination, robot and compute fees, and a citizenās dividend. AI token costs may need to fall 90%: if enterprises are already hunting for cheaper open-weight models, pricing is not a footnote. It is adoption strategy. AIās $3T ROI question: the infrastructure buildout only works if useful AI revenue catches up to the power plants, chips, and data centers being built for it. AI labs are moving up the stack: the model layer keeps getting commoditized, so frontier labs are trying to own the workflows, data, contracts, and habits around the model. The bandwidth tax: the memory story is spreading beyond scarce HBM into packaging, base-die logic, foundries, CXL controllers, and system integration.
| | |
 | We read broadly, and refresh X dot com. Like, A LOT. Pretty telling that Marky Z of FB is back on there as of todayā¦ itās just where AI news happens |
| | | Thatās all for now. | | What'd you think of today's email? | |
|
| P.S: Before you goā¦ have you subscribed to our YouTube Channel? If not, can you? | Love robots? We also just launched a robotics newsletter! Sign up for it here. | P.P.S: Love the newsletter, but only want to get it once per week? Donāt unsubscribeāupdate your preferences here. |
|
