Jeremy Corman

I've channeled 13 years of supporting startup founders with the Digital Marketing Canvas into this innovative GPT. Now, it's in your hands – dive in and make the most of it! ⤵️ chat.openai.com/g/g-RM1hV6bWr-…

Primeras pruebas de Seedance 2.0 desde @chatcutapp Está bastante capado, pero te echas unas risas

@bbalfour @vxanand I agree. But in an AI world, the data you truly own goes far beyond names and email addresses. That kind of surface-level data is quickly becoming commoditized. The real edge comes from truly proprietary data.

I'm a believer in a lot of what @vxanand says here. Marketing in an AI world has flipped. Old World: Channels -> Data New World: Date -> Channels In Pre-AI world most teams first thought through things channel first. What are my channels? What campaigns should I run? What tools (email automation, content, etc) do I buy? Then, they'd try to connect data from other systems like their CRM, product analytics, etc into those tools. This was always a data mess in the channel tools. But it was a tolerable problem. In an AI world, if you want to do anything with AI downstream in your channels it has to start with a good data foundation. So you need tools and people who are not just molding your data, but acquiring new data, combining it in new and unique ways, etc which then enables what you can do in your channels. That is more of an "engineering" mindset and capability. It's why I think Clay will eat a lot of the marketing landscape. If you own where the data is shaped, you can then layer on additional products to execute on that data in various channels. They've been nailing this lately with the launch of Sequencer (email) and their Ads add ons.

@bbalfour If everything can be copied, what’s truly hard to replicate? Not features, but moats: proprietary data, network effects, deep integrations, distribution, brand, switching costs, community, regulation. When shipping is cheap, defensibility wins.

"Coding will be practically solved for everyone within six months." That's the most important line in this talk. It will continue to unleash unbelievable acceleration in what people can build. But it's going to lead to tons of AI Product Debt. Historically, product teams shipped almost everything they spent the hard time building. But that is just going to create a ton of product surface area that few customers use. That creates maintenance, code complexity, edge cases, user support, user adoption bottlenecks, and a lot more. What product teams SHOULD be doing is building early versions of as many ideas as you can. Everyone is talking about "taste." But humans are terrible at judging another person's idea when it's just words. Build it, use it, put it in front of a few customers, send it to 10 people with @reforge Research's AI Interviewer. BUT, kill 9 out of 10 things you build. We've just spent a couple years where incumbents got growth out of shipping a higher volume of stuff. But that's done IMO.

Claude in Excel is now available on Pro plans. Claude now accepts multiple files via drag and drop, avoids overwriting your existing cells, and handles longer sessions with auto compaction. Get started: claude.com/claude-in-excel

In the coming weeks, we plan to start testing ads in ChatGPT free and Go tiers. We’re sharing our principles early on how we’ll approach ads–guided by putting user trust and transparency first as we work to make AI accessible to everyone. What matters most: - Responses in ChatGPT will not be influenced by ads. - Ads are always separate and clearly labeled. - Your conversations are private from advertisers. - Plus, Pro, Business, and Enterprise tiers will not have ads.

To preserve chain-of-thought (CoT) monitorability, we must be able to measure it. We built a framework + evaluation suite to measure CoT monitorability — 13 evaluations across 24 environments — so that we can actually tell when models verbalize targeted aspects of their internal reasoning. openai.com/index/evaluati…

Helping students come up with good startup ideas is like hooking them up to the deck catapult of an aircraft carrier. If you succeed, they're gone. What are they supposed to do, not work on the idea? And working on a startup is incompatible with being a student.

Has anyone experimented with “hidden prompts” in website content for ChatGPT Atlas? I’m wondering if we can guide how Atlas reasons about a page (summaries, form help, etc.) via hidden metadata / DOM hints. Curious about experiments, wins, fails 👀 #ChatGPTAtlas #AIUX

We bootstrapped @TallyForms to $4 million ARR with a tiny team and simplicity as our moat. Here's how we got here: 1️⃣ Pick a crowded, boring industry with low NPS 2️⃣ Target a specific audience (and ignore the rest) 3️⃣ Talk to users constantly 4️⃣ Build one thing extremely well 5️⃣ Keep. Things. Simple. Endlessly grateful to everyone who's helping us grow. On our terms, at our pace, with our values intact. As always, full story on the blog → blog.tally.so/how-we-grew-ta…

@HeyGen_Official Here's the guide for those looking for it: youtube.com/watch?v=yFhTXL…

Your story, told your way. With Veo 3.1 on HeyGen, you can now create multi-scene videos where your identity stays perfectly consistent from appearance to voice. Every transition is seamless. Retweet this for the full guide to Veo 3.1 on HeyGen.

@LottielabHQ Burp

Introducing Burp - Your AI motion editor. ✨ Waitlist is now officially open. Comment 'burp' to be fast tracked on the waitlist and get unlimited credits to test!

It's official. It's genocide.

gpt-oss is out! we made an open model that performs at the level of o4-mini and runs on a high-end laptop (WTF!!) (and a smaller one that runs on a phone). super proud of the team; big triumph of technology.

New Anthropic research: Persona vectors. Language models sometimes go haywire and slip into weird and unsettling personas. Why? In a new paper, we find “persona vectors"—neural activity patterns controlling traits like evil, sycophancy, or hallucination.

ChatGPT can now connect to more internal sources & pull in real-time context—keeping existing user-level permissions. Connectors available in deep research for Plus & Pro users (excl. EEA, CH, UK) and Team, Enterprise & Edu users: Outlook Teams Google Drive Gmail Linear & more

@DruRly Thank you!

@JeremC Perhaps the comments here can help: huggingface.co/spaces/open-ll…

Open-source AI is wild. - WizardLM is catching up with GPT-4-Turbo and Claude 3 - It's not even in the top 5 of the Open LLM Leaderboard Here's what that means for local AI and you 🧵

@DruRly Great thread. Btw, do you know of any leaderboards that rank open-source LLMs?

Want to be in the next report? Look out for the next tweet where we ask for examples.

@levie Remember what ChatGPT did to writing jobs 2 years ago.

Here’s how this plays out: AI makes producing amazing designs more affordable and faster. Designers get to offload the long tail work, iterate faster, or serve more clients or projects. Customers get better results faster or cheaper, inducing more demand for great design.