Tom

CTO at Finlay | Building AI agents for hiring | Proud Malinois owner

[Tom @tomas_builds]

startup X is mostly SF. here's the view from Central & Eastern Europe. I'm Tom, based in Prague. spent the last 4 years in early-stage startups, now CTO at one. the first one failed, hoping round two works. the next one will be my own. building here is different. great engineers, smaller ecosystem, less money, and the local market is tiny so you go global from day one.

i'll post how we build, how a small team uses AI, the wins and the mess.building in CEE too? say hi. if not, follow along.

[Jakub Jurovych @jakubjurovych]

What’s the best model for proactive agents?

Today, we’re introducing the Agent Initiative Benchmark. We tested 12 frontier models on realistic workplace tasks inside a synthetic company with 6 years of highly detailed history: it has a financial model, CRM records, past-due invoices, Linear tickets, and real-life events affecting company development - competitor product announcements, SVB collapse, etc.

We used the company’s Slack to test how a proactive agent listening in on each conversation would assist the team in day-to-day tasks: from stepping in to align initiatives across teams to warning the team not to use an unapproved customer logo.

Highlights:

- 6.1 Sol and Astra from @OpenAIDevs beat Anthropic's latest Fable 5.2, with Sol crushing it at both performance and cost ($0.11 per task vs $1.87).
- The most expensive models aren't the best: Fable 5.1 is 430x more expensive than GPT-6 Luna, with similar performance.
- Most models try to help even if they're not needed. Interestingly, in these failure cases, they spend more time thinking, and the unnecessary work takes about 80% more time and money than if they stayed quiet.

If you’re building agents, keep reading for more tips on which model is best for your use case. Detailed breakdown, the open-source dataset, and benchmark code below:

[Tom @tomas_builds]

@bot is the most impressive AI product I've ever used.

[Tom @tomas_builds]

@danellisona He's selling a course, and he hopes you guys will ask for tips. Wait, you do that.

[Tom @tomas_builds]

@DanielSmidstrup Isn't it the same person you need to hire? Haha

[Tom @tomas_builds]

Spoke about AI with a few friends who work as software engineer in more traditional companies (banks, automotive). They’re living in Dec 2025 in terms of their AI workflows. I felt I'm coming from future.

[Tom @tomas_builds]

@stijnnoorman That’s a great advise. I heard similar one from a founder: “You must be naive, and don’t overthink things”.

[Tom @tomas_builds]

@gregisenberg Doing exactly this for recruiting.

Finlay is an AI recruiter: agents that handle the grind of hiring so small teams don't need a recruiter or a messy ATS. We launched self-serve last week.

Hardest part isn't the AI. It's earning trust in one niche.

[Tom @tomas_builds]

@sankar2704 100%. Give every agent its own scoped identity and the least access it needs. An agent running on the founder's admin token is one prompt injection away from a very bad day.

[Tom @tomas_builds]

@Musty_hasheedu Payment isn't the hard part, trust is. I'd let an agent compare and pick, but the final buy should stay a one-tap human approval until agents stop making dumb mistakes.

[Tom @tomas_builds]

@Naziahasan42 Team of specialists, every time. One generalist agent drifts on long tasks. Small agents with narrow jobs are easier to test and debug. The hard part is the handoffs between them.

[Tom @tomas_builds]

Tried LinkedIn paid ads for our AI recruiter. Lots of visits, no buyers.

Founders who sell B2B: which channel actually got you paying users? Not likes, paying users.

[Tom @tomas_builds]

@alexcooldev How much did the video cost?

[Tom @tomas_builds]

@Microsoft , 2000: "Developers! Developers! Developers!"

Microsoft, today, with Copilot down: "DEVELOPERS?! ANY DEVELOPERS?! HELLO?!"

[Microsoft @Microsoft]

We are deeply sorry for the disruption this incident has caused our customers and partners. We know the trust they place in our services and take that responsibility extremely seriously. We are seeing broad recovery and stability across our services.

This was a technical issue not related to cyber or AI / agents. The incident was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change. We have addressed the underlying issue and continue to restore service to remaining customers and regions.

Given the scale of demand, we restored services gradually, validating stability at each stage. While this took longer, it helped ensure a stable restoration.

We will provide a full post-incident report once our review is complete.

[Tom @tomas_builds]

@spalinskipatryk Lifeless? 6 p.m. on a Wednesday, almost winter, raining, and nobody is out on the streets. Hradec is not Prague, man.

[Tom @tomas_builds]

@Tee_Gee_crypto Thank you, and interested in following what you're building.

[Tom @tomas_builds]

startup X is mostly SF. here's the view from Central & Eastern Europe. I'm Tom, based in Prague. spent the last 4 years in early-stage startups, now CTO at one. the first one failed, hoping round two works. the next one will be my own. building here is different. great engineers, smaller ecosystem, less money, and the local market is tiny so you go global from day one.

i'll post how we build, how a small team uses AI, the wins and the mess.building in CEE too? say hi. if not, follow along.

[Tom @tomas_builds]

@BattMatter thanks! it's going pretty good! we released a self-service product the last week.

[Tom @tomas_builds]

@inferredbylisa Take care, hope they left you some equity.

[Tom @tomas_builds]

@yasser_elsaid_ She succeeded in her mission, destabilising the US's most valuable company. She asked for the promotion.