Noah D. Mitsuhashi

Still here, working on itsdoneai.com Travel isn't solved. I don't want to define my own preferences, look up emails and calendar events, etc. Too complex for a wrapper.

@conor_ai I don't totally understand how this stops prompt injection, couldn't the outie relay the bad request to the innie?

agents can and will share sensitive info with anyone they’re talking to if you share private thoughts about a customer, colleague, or manager with an agent, it won’t hold back from telling them directly context engineering for agents means heavily constraining both who they can interact with and what data they have access to assume anything an agent can access will be shared with anyone it can reach for @hyperspell we split this into an ‘innie’ and an ‘outie’ agent the innie has access to all internal data but can only talk to our team the outie can talk to customers but is heavily restricted in what it can access

@NoamTenne @claudeai Gambling as productivity tools

@noahmits @claudeai Productivity tools as rollercoasters

I used to be a total @claudeai boi and I'm sorry but this is now true

@pmarca Reject modernity, embrace Haiku 4.5

Frontier models become more expensive to serve. Token demand explodes at exponential rates. Solve for the equilibrium.

Magical OpenClaw experiences that use frontier models cost $300-1,000/day today, heading to $10,000/day and more. The future shape of the entire technology industry will be how to drive that to $20/month.

My moat is that other companies don't have me as the founder

@vikhyatk I respect it but you shouldn't have to cancel it yourself itsdoneai.com

> wake up > flight takes off in 2 mins > open amex app > cancel and get full refund this is why america is the greatest country on earth

@Govindtwtt "Book <friend>'s wedding" actually finds me the flights and hotels I'd pick myself and organizes it around my schedule itsdoneai.com

we are being gaslit about AI on a societal level. Everybody is vibe coding but I haven’t seen one useful thing get produced. Everybody has agents doing something but nothing useful is getting done. Cool you had AI summarize a PDF and make a template. Nice

@geoffreywoo Travel should only be three steps "Book my family vacation to France" Yes those flights and hotels look good Book itsdoneai.com

psa to every ai startup founder: your agent isnt autonomous if it needs human approval every 3 seconds real agents make decisions, execute trades, hire contractors, and deploy code without asking permission if your "ai agent" sends slack notifications for approval, you built a really expensive chatbot

@pacificleo It needs to be easier/more powerful than doing the work yourself, people like using ChatGPT for planning because it's powerful but the actual booking there is slower than just buying it yourself Needs to be free and easy af "Book Coachella" itsdoneai.com

Nobody I know has made a hotel booking or bought an airline ticket through the agent. Why do all these companies use that use case as an example while talking about agentic capabilities?

@DietCoke_Esq It's doing it's best, but we think we can do better itsdoneai.com

Yes Google Flights, I don’t mind a 21 hr 10 min layover between Boston and Dallas. Thank you for the recommendation

@sabakarimm @UberEats The last item lets you manually edit it

Dark pattern alert: @UberEats doesn’t allow going back after going through six step order process. It’s add tip or lose your whole order.

@thekitze Or itsdoneai.com

this would be a crazy concept but until then u def need fLiGhTy

@luvvraq You're gonna love us Mobile first "Book Coachella" itsdoneai.com

i paid with apple pay too idgaf

@coreyganim Considered it then decided to do the whole thing end to end. Proactive travel planning, organizes trips for you, decides based on preferences etc. you shouldn't need to pick your own flights and hotels you should just say "Book my meeting next week" itsdoneai.com

the business nobody is building with this: 1. pick a niche that travels (consultants, sales teams, remote agencies) 2. build a travel monitoring agent that watches their routes daily 3. when prices drop below a threshold, it auto-alerts them or books it 4. add hotel + car rental monitoring on top 5. charge $200-500/mo per client as "travel cost optimization" a 10-person sales team spending $50K+/yr on flights would pay you $500/mo without thinking twice if you save them 20%. one API. one agent. recurring revenue.

@brianwilt I'm pretty excited for ours itsdoneai.com

Which has the more useful info, the Google App or Apple Wallet haha

@rdominguezibar We turn "book my meeting next week, I want to go to an escape room" into the flight hotels etc. you'd book yourself itsdoneai.com

I'd love to angel invest in a handful of startups this month. Pre-seed and seed. Ideally AI or VC-adjacent, but open to all. My value add: ▪️ 500K+ newsletter subscribers across The VC Corner and The AI Corner ▪️ 300K+ LinkedIn followers, 2–4M weekly impressions ▪️ a16z speedrun scout ▪️ Network of top VCs, operators and founders 👉 Pitch in comments

@andrewchen This has been our main focus for making Travel feasible. Single digit cents of tokens to plan exponentially more complicated trips. I've seen a lot of powerful tools that are extremely vulnerable to token prices and scale exponentially with more complicated trips let alone groups

for AI-native consumer apps to be truly ubiquitous we need: ARPU > Average Inference Cost Per User. How far away are we from that? Ideally AI native apps can hit APIs on every screen yet also pay for just by throwing some ads on it But I think we’re a while away: - unfort we’re probably >10x off right now. Monthly ARPU is $2-5? Token costs for an AI heavy app might be $20-50 of cost - much of consumer is global. Even if we hit the US/EU it’ll be a while before we can serve the broad base - more importantly every time AI improves, consumers demand more. No one wants to talk to a last gen AI character. If video gets good they’ll want videos everywhere - same issue for LLMs. We started with short replies back. Now we want agents who can do entire tasks for us - it might be we need major innovations in small models or new mobile hardware so that we have free local inference. But that will still not be as good as SOTA cloud LLMs obv, but might solve some use cases No wonder so many products focus on productivity, and on prosumers who can pay $100s or $1000s on work related tasks. This is where you can have huge ARPUs and benefit from being SOTA. And it seems as though there’s no limit for tokens… so why do the low end?

@jeiting I think there's a reasonable number of verticals where paid acquisition is the norm, namely, low or no switching costs Travel, Health, CPG, Beauty, Retail, Entertainment Navan, Blade, JSX, Ro, Hims, ~Ozempic, Instacart, ~Food Delivery, etc.

Olivia’s being blown up here but she’s mostly right but maybe misunderstood. “Generational startup” is a word thrown around a lot in VC land but generally means “billion dollar plus valuation and very durable”. I.e. the outcomes VCs generally look for, these are the bets that make funds work. These outcomes (Duolingo, Roblox, Netflix, Spotify), are extremely rare, and only ever get there because they have a core PMF so strong the market pulls them to scale. They will use ads to supplement their growth, but only as an accelerant to an already compounding product and brand. Usually driven by word of mouth, which is a side effect of PMF. Apps whose primary driver of growth is paid ads, can reach great outcomes and even considerable scale, but the dynamics of paid acquisition mean they are less durable, and less capital efficient. Not bad companies, but maybe relatively “bad” investments. They are just too easily disrupted without constant acquisition knife fights. So, just know the game you’re playing, have the right capital structure, and most paths can “work out” for some definition of “work out”.

@omooretweets I think there's a reasonable number of verticals where paid acquisition is the norm, namely, low or no switching costs Travel, Health, CPG, Beauty, Retail, Entertainment Navan, Blade, JSX, Ro, Hims, ~Ozempic, Instacart, ~Food Delivery, etc.

There are of course some exceptions - ex for a content graph driven product (TikTok), you need to pay for early density And some might argue that with users paying more for consumer AI products, you don’t need to have a TAM of “everyone” But I’m still skeptical!

In consumer, paid ads generally = lack of true product market fit I have yet to see a generational startup with largely paid ad-driven growth…