.@NataliaZarina rode so hard for Claude Code we devoted an episode to how she was using it to automate her job running @every’s consulting practice.
Fast forward to five months later, and she rides just as hard for Codex. I had her back on AI & I to talk about what caused her to make the switch, including how she ran a prompt in Codex before bed and woke up to a finished, custom CRM tool.
We get into:
- Why she finds Codex easier to use than Claude Code
- How she’s using loops in Codex to create customized tools that work exactly how she needs them to
- Why the consulting team still pays for SaaS products like Attio and Asana even though they could vibe code their own versions
- How she built an app to manage her father’s medical care in Codex
- How knowledge work is evolving from sculpting to gardening, in which you develop the context and logic you need for an agent to execute for you
This is a must-watch for anyone trying to figure out whether to build their own tools or buy real software—and what it takes to get an AI agent to run unsupervised for hours and nail the output.
Watch below!
Timestamps
1. Introduction: 00:01:05
2. How Natalia manages Claudie, the consulting team’s AI project manager: 00:02:35
3. Why the consulting team still pays for SaaS products: 00:04:55
4. Codex as a game changer : 00:11:47
5. Building personalized learning guides and illustrated explainers with AI: 00:14:55
6. Inside Natalia's AI-powered email triage system: 00:21:40
7. The shift from knowledge work as sculpting to knowledge work as gardening: 00:26:44
8. Using Codex to on-shot a custom CRM: 00:28:57
9. Using Codex to build an app that coordinates her father’s medical care: 00:33:16
Somewhere in your 20s or 30s you’ll get the opportunity to rebuild your life after a negative loop, heal from what broke you, live in your own space, reconnect with your discipline, and learn to love yourself again. It’s very important that you see that journey through.
YOU MUST BECOME MORE CONFIDENT
Be egoistical. Dress well. Take care of your body. Auramaxx. Make others feel your presence. Insecurities are in your head. Love yourself first. Don’t be a people pleaser. You’re the priority. Play big. Overthinking drains your energy. Nothing is that serious. Opinions are just opinions. They don’t matter.
From shy → confident.
From insecure → empowered.
From scared → brave.
When you know yourself and work on yourself, nobody can make you feel small.
Look at people you admire.
Do you think they care what others think of them? You think they are not confident in themselves?
The answer is obvious.
The secret to everything is confidence.
Every door you thought was closed, you’ll realize was open the whole time. You were just afraid to step in because of the unknown, the darkness there.
But you forgot that you were the bright light all along, meant to shine.
BEING SHY RUINS YOUR LIFE
You’re losing opportunities. Wasting potential. Low on energy. No drive. Lacking courage. Living in fear. Fix it. Stop caring about other people’s opinions. Get what you want. Be unapologetic. Don’t ask for permission. Get angry about your life. Start
ok I spent this morning inside Riverside's MCP integration with Claude and I'm convinced this is the new meta for video editing
before this, you used to have to edit in Descript, clip in Opus Clip, hire an editor to stitch it all together, and schedule in another tool
but all of that just got swallowed into Riverside.
in 10 mins i was able to:
> edit the video by editing the transcript. delete a sentence of text and that part of the recording disappears with it
> AI pulled the best moments into vertical clips for me, captions and all
> draft my newsletter straight from what I said
> scheduled my posts out across all my social platforms
so the whole record → edit → repurpose → publish loop now happens in the same tab I recorded in
and because of the MCP, I literally just told Claude "clip the 3 best moments from my recording and queue them for the week" and it goes and runs it for you
very based agentic workflow for content
AI Videos are ALL slop.
AI should be making you a content machine.
Introducing Riverside 2.0, the first AI Producer that creates authentic content while you sleep:
Most companies are implementing AI agents RANDOMLY for non-coding use cases... And it's hurting more than it's helping.
Over the past few months, we've helped 50+ companies implement AI agents for ads, content, operations, analytics negotiations +more. Here are 8 useful workflows that have driven the most tangible value.
1] Ad scraping / Competitive & Market Intelligence
A company called “Foreplay” (weird name) has an API that lets you scrape your competitors' ads, sort them by performance, and download the actual creative assets directly. This is extremely useful for ad teams. Instead of asking an agent to write from scratch, you can have it constantly study the best-performing ads in your market, pull the patterns, and use those examples to generate better hooks, scripts, visuals, and campaign ideas. Putting this agent in Slack where the entire marketing team can access is incredibly useful.
2] Organic Content Scraping
All AI models are bad at writing content scripts, ads, and other forms of media... until you ground them in high-quality examples. When you let an agent research any niche, scrape the best-performing posts across platforms, and understand what's actually working, the output gets dramatically better. This can take real setup time because scraping services are often fickle, but once the data pipeline works, it becomes one of the highest leverage inputs for content/marketing. For those making videos this skill is really useful because agents can download full youtube videos, and cut relevant parts of the videos down and organize them into folders. Video editors LOVE this one for B-Roll and cinematic edits.
3] Content Re-Optimization
Listen to me: AI agents can watch your videos (Gemini has great video analysis abilities) and make precise edits if you give them the right tools. So many marketers make a video, it underperforms, and then they just go make an entirely new video. That is usually the wrong move. A lot of the time, the video does not need to be redone... It needs 1-3 micro-optimizations:
> better / shorter hook
> cleaner cuts
> better first frame visual
> stronger caption
When you set this up properly, an agent can find the weak parts of the video, make the edits, and even re-upload the video or rerun the ad automatically.
4] Make All Meeting Notes ACTUALLY retrievable (With Permissions)
Almost everyone has AI joining meetings to take notes. But very few companies have a system that keeps those notes organized, searchable, and permissioned properly. Anyone on the team should be able to find important context from past conversations in under a minute. You should be able to ask the company memory, “What did we decide about this?” and get the answer instantly, only from the conversations you’re allowed to access. Amazing how many companies just record stuff and people have no idea how to ask a question about that data.
5] Automated HIGH SIGNAL Dashboards
Most companies produce too many summaries, too many digests, and too many random links with too much information. Many companies we've talked to actually get annoyed with too much information. One of the highest-leverage uses of agents is deciding the FEW metrics that actually matter, organizing them into a clean single glance dashboard, and having an agent send a concise daily summary of what changed and what it means. The agent's job should be to decide what actually matters. This can't be achieved without a good amount of upfront work. It's worth it.
6] Lifecycle Email Agents
Tools like @loops and @resend give agents control over sending, sequencing, and personalizing emails. Instead of one static onboarding flow, an agent can manage the full customer lifecycle: welcome emails, activation nudges, abandoned signup follow ups, renewals, win-backs, etc. This one takes time to set up, and I still think it should be done with human in the loop, but it saves SO much time, especially if you configure skills properly.
7] Brand Negotiations
We've been setting up agents for creator agencies that manage large creators on Instagram and TikTok. With Fable level models, agents can handle 95%+ of the negotiation process... collecting campaign details, comparing terms, drafting email replies, flagging risky clauses, and escalating the parts a human should approve. I think agents will handle most of the back-and-forth over email, as long as the human-in-the-loop checkpoints are extremely clear before anything is agreed to or signed.
8] Genmedia
@fal has a “genmedia CLI” that can use any creative model (Image + Video). When you configure it properly, and only route people to the best models, it becomes extremely useful for marketing teams. You can put it directly in Slack so the whole team can generate ad concepts, content assets, website visuals, product images, and creative variations without jumping between a bunch of different external tools.
These are all things we've noticed when working with companies helping them become more @agentnative_. Videos on all of these individual workflows soon!
everything you are spending a lot of time on should be run through the basic audit
ask yourself:
>does it make me smarter
>does it make me healiter (+ mental health)
>does it make me money
are all of the answers negative? you are wasting your life
Anthropic engineer:
"You can build 5 assistants in one afternoon. Each one handles a task you've been doing manually every single day."
In 45 minutes he shows exactly how to do it from scratch, step by step.
Most people are still doing all of this by hand.
Watch the session, then save the guide below.
If you want to understand the difference between human-created art and AI-generated slop, watch this talk.
It's about the design process behind Charli XCX's cover art for "brat", but the designer spends the first 20 min telling his childhood story.
It feels random at first, until you realize he could not have created what he created without those lived experiences and obsessions.
Art is the culmination of one's lived experiences, whereas an LLM's output is a statistical average of its training data.
"As a designer, our real work is not in the actual design itself, but in collecting ideas, being inspired, seeing other things and taking it into your brain. That's the work.
The work is not spending time on the computer trying to get the thing right. THE WORK IS EXPERIENCING, taking it all in, and consuming media and content to the craziest degree."
(highly recommend watching the whole thing, link below)
If I learned one thing about job searching in 2026, it's that you have to make your own side doors.
So let me try something new.
If you are (a) familiar with my work, and (b) close to someone at @GoogleDeepMind, please slide into my DMs!
→ Do a 24-hour social media detox
→ Photograph textures around you
→ Memorize a short poem
→ Learn to make a perfect sauce
→ Sit in silence for 10 minutes
→ Attend a public talk or an open mic
→ Volunteer once without posting it on social media
→ Build something with your own hands
→ Fix something in your home
→ Watch a classic movie from beginning to end
781 Followers 7K FollowingHyperdidactist.
Philologist; Obstinated-Criticist.
Support Ecology-Economy of Minds
PhD ENGINEER {semantic,graphic}memory
IN SCIENCE We Trust. #TheoryWave
6 Followers 24 Following8 yrs in B2B tech. went from clueless junior to leading a team. sharing what i learned about promotions, negotiation, and not losing your mind at work.
1K Followers 7K FollowingLiving life one delicious bite at a time 🍽️
Exploring flavors, savoring moments
Food lover on a tasty adventure 🌍
#FoodieLife #DailyEats
1K Followers 1K FollowingHistory • Geography • Science.
Telling the story of our planet through time and space 🌍🕰️ • I do not tolerate internet bullies.
311 Followers 3K FollowingGTM and AI marketing adventurer. I do also build apps and open source them.
Currently: AI that runs on your Mac with no internet.
56K Followers 115 FollowingSimplifying AI for non-technical entrepreneurs. Built an eCom biz to $16 M+ revenue. Built (then sold) a coaching biz to $340 k ARR.
32K Followers 5K FollowingDreaming about the future, romanticizing the past | MP @compoundvc investing & researching former science projects & crypto | we are only creativity constrained
43K Followers 3K FollowingFounder https://t.co/IPScOqNfhr. I build software for meme marketing and memetic warfare. Taking silly memes deadly seriously.
5K Followers 963 FollowingDark Empress of Fintwit. Writer. Geopolitics, philosophy, tech, finance & more. American formerly lost in Paris & Amsterdam. Ripostes/reposts =/= endorsement
1K Followers 636 FollowingStarted vibe coding at 49 with zero skills. Now shipping AI after kids sleep. If I can do it, you can too. 🇮🇹 Dad of 2 https://t.co/9eSFGy8er5
3K Followers 1 Followingwe're making AI models & agents feel right, starting with design.
post training data | RLE | online & offline evals | context tools for agents
6K Followers 546 FollowingCreate a life too fulfilling to numb •
Exec coach + somatic guide •
Men’s work + retreats: https://t.co/NqD1u2xwnu •
Writing: https://t.co/vrLIV2STBD
36K Followers 2K FollowingDesigning at @IterationDesign, teaching at https://t.co/Jn9brli34Y. Previously led teams at Dropbox, Wealthfront, Groupon & more.
8K Followers 920 FollowingBuilding things for people who build things. Design Engineer @NotionHQ. Previously @GitHubNext. Creator of https://t.co/uuE5gts5MF. Working on @lumen_notes
104K Followers 3K Followingcreating something new. product builder, most recently at meta. my last startup was acquired by twitter so this place has a special spot in my heart. 💙
2K Followers 2K Followingwannabe cat, curator of taste and designer of brands ♡
works traceable at @DacoitDesign @exaprotocol @shopflo_, c7 @10kdesigners
614K Followers 1K FollowingWe make very small computers which you can buy from just $4. We are also the literal coolest. Be excellent to each other. Tech support: https://t.co/ZEBSfmuErK
270K Followers 1K FollowingFounder @teamSundial. Angel investor. Author of "The Making of a Manager" https://t.co/6HwJhCW5Hi. Obsessed with systems. Design + data person.
4K Followers 943 Followingcertified freak, 2 days a week. the other 5, i do marketing things @Lovable. past lives @baggu @claudeai @partiful @perplexity_ai @NotionHQ