Art Stavenka

@garyleff Fair on air (but still need to scale-test). Other categories like hotels are different, no SITI equivalent. OTAs price per POS, gaps stay wide for long. 84% of our 1k hotel sample had a lower price elsewhere. The pain is invisible, you pay $400 never aware it's $328 elsewhere

@stavenka Country-specific fares (SITI) exist but are not dominant. They will simply disappear if they no longer let airlines price discriminate. The purchase process for tickets is not the primary consumer paint point in air travel.

Even A Nobel Economist Got Stumped By Airport Self-Check-In — Travelers Still Need A Human [Roundup] - View from the Wing viewfromthewing.com/even-a-nobel-e…

@kathrynwu1 There's a flip side - clear writers learn to code very fast, much faster than the reverse. Two of the best engineers I've worked with came from journalism and law. The thing being trained is the same precision muscle, just from different starting points

I think one reason YC likes logical engineers is not just because they can code. A lot of them are unusually clear communicators. Coding trains you to think in strict logical sequences: input → output, cause → effect, constraint → solution. You can hear it immediately in good founders. Not necessarily charismatic, but coherent. People underestimate how much startup momentum comes from simply being easy to understand.

@skift Travel chatbots in 2026 mostly answer 'when is my flight' and call themselves agents. Tara will be different the moment it can actually rebook a hotel at a price the user couldn't find. Until then it's an IVR with a personality

ixigo relaunched its app with Tara, a voice-native AI assistant, targeting India's mass-market travelers who prefer voice interaction in multiple local languages. India's unique linguistic diversity, high adoption of voice technology, and the dominance of Android have made it a testing ground for voice-first travel solutions. A robust open-source voice AI ecosystem is emerging in India, setting a reference design for travel tech that is likely to influence global markets. hubs.li/Q04hRVmd0

@garrytan This is exactly the loop. We run 1M agent searches/mo, log every disagreement between models, the eval becomes its own training corpus. The dirty truth is the eval data is more valuable than the prompts

Funny how simple using openclaw and Hermes agent is these days Just have it do stuff. Then improve in progressive batches with evals from multiple frontier models. It self improves!

@Suhail People think AI company = foundation model or chat. The real expansion is post-LLM like agents taking actions inside someone's billing flow, supply chain, booking funnel. That's where the next 1000 companies are hiding

Possibly the thing we will most realize looking back: intelligence was so big that lots of companies were going to succeed. It's not so simply bucketed into chatgpt and claude code.

@levie Built an AI agent that's right 95% of the time on the demo flow. Took us 6 mo to get the last 5% production-safe across 100 countries and edge cases the demo never hits. Point is the CEO demo and the prod system are different products

CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI. So when they play with AI, they see the happy path results, often not considering the next 10 or 20 things that have to happen to get sustainable results from agents. “Look I made this awesome product prototype”. Yes but you didn’t have to review the code before it went into production and fix a bunch of issues. “Look I generated a contract”. Yes but you didn’t verify all the terms before it goes out to the counterparty and didn’t have to wire up all the past contracts to work with. The best thing you can do as a CEO is to use AI a *ton* to figure out the real implications of agents in the enterprise, and come out the other side with an appreciation for both the upside and the real work that goes into them.

@HarryStebbings The Anthropic-doubled-and-we-wouldn't-blink one is the real tell. Compute is now cheaper than the SaaS subscription tier you'd negotiate down to. CEOs who get that are quietly rebuilding their stack. The rest are still on the renewal call

I just interviewed a CEO who said three things that blew my mind: 1. We replaced our $600K Salesforce contract with a vibe-coded CRM, built within 3 weeks. 2. We will get rid of 80% of the SaaS we use internally. 3. If Anthropic doubled pricing, we would not change usage in any way.

Live in Canada? You throw away $294/year on travel. Just because of your country. UAE $336. France $168. The hidden charge people can't see. When you book travel, lowest price is hidden in one of other countries. And AI finally got good enough to find it & give the money to you

@garrytan Skillify + cron is the boring 20% though. The fun part is when your eval suite passes and the agent still does something psychotic in prod because the world shifted under it overnight. Hotel prices, currencies, AB tests. Build for that or it falls over

This is exactly how I build my agents and I’ve built 4 back to back now: both personal AI and company brains It’s the same methods over and over. Do it. Skillify it. Add to cron. Check if it is resolvable. Evals and integration tests. Repeat.

@snowmaker LOC is the only metric where this comparison flatters AI. A 450-eng team gets you architecture decisions, tradeoff arguments, weird customer support escalations resolved at 2am. $2M in tokens gets you a lot of confidently-wrong refactors. Different units probably

I did the math, and $2M in OpenAI tokens is equivalent to a 450 pre-AI human engineering team for 2 years, in terms of raw LOC output.

@fin465 12 meetings/day for 3 days = 36 meetings, maybe 4 convert. Same effort spent on 30 thoughtful comments under the right founders' tweets gets you 30 inbound DMs from the actual ICP. Conferences are one sales channel dressed up as a GTM strategy

talked to a YC company that scaled from $0 → $2m ARR in their first 6 months with their ENTIRE GTM built off going to conferences. Here's the playbook they cracked (step by step): ~4 weeks before: > Post abt the conference and tell attendees exactly how to reach you > Send personal DMs to the right ppl on LinkedIn and X > Reply within the hour & lock in 10 top targets to close. > Send everyone else to your drip email campaign. Then, set a meeting block of 1-3 days during the conference: > make shared booking link for the team > Reserve a quiet café / private dining room > Pack in 12 meetings per day, 30 min each, with buffer time built in While you're there: >Hand every prospect a thoughtful small gift and a personal card >Single out 5 standout customers whose pain ur product actually solves >Pull them aside for a casual on-camera Q&A in a solid film spot >Don't pitch hard. >Let the conversation breathe and weave your product in naturally. The 4 weeks after >Hand the raw footage to a freelance editor + ask for ~15-20 punchy clips with captions. >Drop a new clip every couple of days on LI / X > use these clips when you post online about the next conference to keep the momentum This is the formula, costs less than a few thousand dollars to execute. They’re on track to end the Y1 at ~$6m ARR (B2B, targeting large enterprises) + STILL not using any other channels for customer acquisition

@sweatystartup The split is generation vs. execution. We run 1M agentic price searches per month for travel. Nobody reviews say hotel bookings. The work that used to be 'check 6 sites' becomes a refund hitting your card. Zero downstream slop. AI as filler is the bad case

The people using AI really heavy think they are 100x more productive. In reality they are a bit more productive but making 10x more work for other people. More contract comments. More bullets. More observations. Longer interview processes. Longer emails. More asks. More distraction and more work. They think they're getting more work done. In reality you are driving your employees, coworkers, partners, vendors nuts.

@levelsio @Airbnb Airbnb went from couchsurfing-with-stakes to a hotel chain that legally can't say it's a hotel chain. Once unit-level reviews mattered to the hosts' income more than to the guests' info, the platform was always going to side with the host

Same, I really don't like @Airbnb after booking my last one in Brazil I used the shower for 3 minutes and it was already flooding, so I told the host, who started blaming me for it Then contacted Airbnb Support, who called us etc. Then in the end the host gave me a bad review with fake made up shit, and Airbnb wouldn't let us post a review about the host! Just speechless really

@skift Agentic commerce is going to standardize the part of the funnel that already mostly works. The interesting question is whether agents start optimizing AFTER the purchase too because every booking is mispriced relative to some other IP. Different problem

Google on Tuesday named hotel booking as the next vertical for its Universal Commerce Protocol, publicly committing to extending its agentic commerce infrastructure to travel. The announcement came via a blog post by Vidhya Srinivasan, vice president and general manager of ads and commerce, alongside a wave of shopping updates at the company’s annual I/O developer conference. UCP, the system that lets AI agents handle purchases inside Google’s search and chat interfaces, is expanding “to even more verticals, starting soon with hotel booking and local food delivery,” Srinivasan wrote. hubs.li/Q04hLbyl0