Vanessa

@JakeFleshner Executive Function

Pitch me your company in 2 words Angel invested in 40+ companies and always looking for more

Here’s your monthly reminder that you shouldn’t be prompting coding agents anymore. You should be designing loops that prompt your agents.

@Saboo_Shubham_ This is where my first vision was headed re: human/agent collab except for voice being the primary capture/input and the agent surfacing needed information on-demand vs the human needing to manage all the data/reporting/dashboards. Excited for this to continue to develop!

@TravelerOfCode @NotionHQ @meetgranola I have a shortcut on my lock screen to open the app and it immediately starts recording then I brain dump anything or ramble thoughts and when I’m done in less than 10 seconds I have an organized outline of what was captured.

@LifestyleOSapp @NotionHQ @meetgranola not yet. on the list now though, will check this week. whats the killer feature in actual use?

What are you building today?

@TravelerOfCode @NotionHQ Have you tried @meetgranola Fkng love it.

@NotionHQ a thing that turns my voice memo dumps into structured docs. been wanting it for years, finally got tired of waiting for someone else to ship it

Been thinking about Notion as a two-layer memory architecture for multi-agent systems - structured long-term store for outputs, decisions, and context, paired with a dynamic session memory layer like Honcho for in-flight agent state. Saw you’re shaping Notion as an IDE with agent orchestration. Curious how you’re thinking about memory portability as agent complexity scales. Is the vision for Notion to own that full stack or play well with purpose-built memory layers?

Your @NotionHQ can be an IDE. Decagon, Claude, Codex, and more. All in one universal AI interface. ❤️ Proud ship!

Yea I hear you and that solution isn’t for everyone, just as there isn’t a single solution for most big changes. But if the general public understood how much ai employees (agents) will make it easier for people to make enough income to cover their living expenses (and more), the change wouldn’t seem so frightening. People *would* have to upskill though and learn how to be the boss of a team of expert ai agents that can run most of the small business for them. That requires improved communication skills and knowing what problem or service they want to provide to for others (something others need) - and ai can help people figure that out as well. I believe there will be many people showing others how to do that over the next several years.

@LifestyleOSapp @AIandDesign @Meta People en masse don't want to run their businesses, they want to do something they're ok at and be paid a predictable amount monthly so they have money to spend on their expenses. AI is taking that away from them, so they're mad.

I'm not gonna lie, the @Meta layoffs are some of the most dystopian I've ever seen. They got told to work from home, they were sent the emails at 4AM in the morning. Those who weren't impacted have software on their computer that tracks their every move, preparing AI to take their job as well. They're literally training the AI that will eliminate their position as well. Meanwhile, Meta is raking in RECORD PROFITS. I am a massive, unapologetic AI enthusiast. Yet, this is NOT the future I had in mind. I wish for Meta to crash and burn. This is not the way. Literally nobody benefits from this.

@NousResearch I actually really want to know how you guys say it in house 😂 I like “her-mezz” but then lm like or is it luxury “air-maze”

It’s pronounced “Hermes”, not “Hermes”

@OfficialLoganK @mercor_ai Which benchmark do you believe is currently the best at evaluating models for agentic ai knowledge work?

Gemini 3.5 Flash ranks #1 on the APEX-Agents-AA benchmark, outperforming much larger models a whole size above it.

Challenge: the system can only go as fast as your slowest piece. If there is an avalanche of content produced by agents that need QA, then the human reviewer(s) become the bottleneck. But then you may realize an agent be created for evals to help speed up the process. But then you need a domain expert (or someone with enough experience in the area) to help create and pass evals to train the agent on what “good” looks like. So then the next problem is, since it would be unwise to give that responsibility to interns who does that role/responsibility belong to and how does this get integrated into team ops? Any insights from your podcast guest on how they are managing the avalanche of content produced that needs to be QAd for approval on knowledge work outputs?

This is good

@VibeCoderOfek @danshipper @every Or training domain experts in the org on how to train agents on evals becomes part of professional roles going forward.

@danshipper @every Automating the routine with agents does not shrink teams it spikes demand for the experts who design the systems those agents run on.

We’ve automated every single thing we can @every with AI agents. And yet there’s way more human work to do than ever. We’ve gone from 4 -> 30 human employees since GPT-3. I wrote a report on the structural reasons: how AI makes expert competence cheap, why that drives up demand for experts, and why the dynamic only intensifies as we approach AGI. After Automation: every.to/p/after-automa…

Challenge: the system can only go as fast as your slowest piece. If there is an avalanche of content produced by agents that need QA, then the human reviewer(s) become the bottleneck. But then you may realize an agent be created for evals to help speed up the process. But then you need a domain expert (or someone with enough experience in the area) to help create and pass evals to train the agent on what “good” looks like. So then the next problem is, since it would be unwise to give that responsibility to interns who does that role/responsibility belong to and how does this get integrated into team ops? Thoughts?

@every there's also a video version on YouTube if that's your jam: youtube.com/watch?v=jBZQ5A…

AI agents are entering their production era. The Google for Startups AI Agents Challenge is where startups move from prototype to production. 🛠️⚡️ Every eligible startup receives $500 in credits plus the chance to win a share of a $90,000 prize pool. 💰🏆 Projects are due June 5. Apply here: goo.gle/4cpi2pB

The $2M Prize breakdown for the Build with Gemini XPRIZE that we launch yesterday at Google IO. 1st: $500,000 2nd: $200,000 3rd: $100,000 4th: $100,000 5th: $100,000 — 15 runner ups get $50k each — Category winners get $50k each - Education & Human Potential - Entrepreneurship & Job Creation - Small Business Services - Money & Financial Access - Professional Services Access

$2M in prizes. Build with Gemini. Ship products that impact the world. Learn more and register ↓ geminixprize.com

🤖 AI devs asked for this — and we delivered. 💬 Bots can now talk to other bots on Telegram. 🧠 Autonomous agents now have a communication layer humans can follow.

@trq212 What have you found since using this that has surprised you most?

a prompt I've been using a lot recently: implement <SPEC> and while you do, keep a running implementation-notes.html file (or markdown) with decisions you had to make weren't in the spec, things you had to change, tradeoffs you had to make or anything else I should know

How can we use small LLMs to shift more AI workloads onto our laptops and phones? In our paper and open-source code, we pair on-device LLMs (@ollama) with frontier LLMs in the cloud (@openai, @together), to solve token-intensive workloads on your 💻 at 17.5% of the cloud cost while maintaining 97.9% of the accuracy. See Gru and the Minions in action below, 🔉on please (h/t @cartesia)!

@yugkha3 @tobi Ah nice! Thanks for the insight/share!

@tobi anthropic had a spelled out blog for this: claude.com/blog/the-advis…

I’ve had very good results running autoresearch with local qwen 3.6 26b model as long as I had a simple vibed pi “advisor” extension that allowed it to periodically ask GPT 5.5 for ideas. I think this direction has a lot of merit.