Dmitry Belaventsev

Write the People, Talk with Code
17 yrs SWE → trading → AI agents
Polystrike founder · building in public

[Dmitry Belaventsev @dizpers]

Got fired in November. Applied through December. Zero interviews.

I have 17 years in software engineering. Last 6 years — fintech and crypto, trading systems, CEX/DEX integrations. Built teams, led projects, stayed hands-on with code the whole time.

Something changed though. It's not the usual November/December hiring freeze — I've been through those before. This time I'm not even getting invitations to interview.

So I stopped waiting and made three bets:

(1) Find alpha in prediction markets. Already built the system, already trading with it.

(2) Learn to build AI agent teams. Same management skills I've been developing for years — delegation, trust boundaries, feedback loops — but the team isn't human.

(3) Ship my own products. Stop depending on someone deciding I'm worth hiring.

I'll share what I learn along the way. First topic — how to give AI agents memory that actually improves over time.

[Dmitry Belaventsev @dizpers]

@weswinder $100 for 5x max
$200 for 20x max
so your option should cost $300 I suppose

[Dmitry Belaventsev @dizpers]

@orlasuture @marclou Those choices are quite common in the industry -- how to render the pages, how to send emails, how to get paid

[Dmitry Belaventsev @dizpers]

@marclou is making $3K/month selling a NextJS boilerplate for $169.

As someone who spent 17 years writing code, my first reaction: why would anyone pay for this in 2026?

Free templates everywhere. AI generates full project scaffolds in minutes. Open source starters cover auth, payments, email.

But the numbers don't lie. 359K followers and $3K/month from a product that bundles what most engineers would set up themselves in a weekend.

Still wrapping my head around it honestly.

[Dmitry Belaventsev @dizpers]

@robj3d3 Looks promising. Tried another tool recently https://t.co/bWYVi2Uq0x

But it doesn’t deliver posts ideas. I mean those posts ideas the platform discovered in your niche

[Dmitry Belaventsev @dizpers]

@Borg_Cryptos 1-1

[Dmitry Belaventsev @dizpers]

@Borg_Cryptos 2-0

[Dmitry Belaventsev @dizpers]

@Darky1k 2-0

[Dmitry Belaventsev @dizpers]

@nicklaunches https://t.co/DmDVDyarZ5

infinite ai generated worlds
NPCs living their lives between your moves
ai text rpg
completely free

[Dmitry Belaventsev @dizpers]

@Joi2James Just try this ai text rpg https://t.co/DmDVDyarZ5
You'll love it

[Dmitry Belaventsev @dizpers]

@sflorimm No

[Dmitry Belaventsev @dizpers]

@jpschroeder maybe $300/mo (like next tier after max 20x)

[Dmitry Belaventsev @dizpers]

@sergeynazarovx ok

[Dmitry Belaventsev @dizpers]

@sergeynazarovx Can’t you just make housekeeping yourself then?

[Dmitry Belaventsev @dizpers]

Three things I'd tell any new Polymarket trader, and not one of them is a pick.

1. Survive first, optimize later.

Sizing beats being right. You can have a real edge and still blow up on variance if you're reckless with position size. Cap it: 2-3% per bet, 20% total at risk on any single day. The 5-10% unit crowd is right about the pick and wrong about the math of ruin. They just haven't hit the bad streak yet.

2. Assume you have no edge.

Liquid markets are priced by people with better models than you. They sit within a cent of the sharp book. "I found an edge on a liquid market" is usually a measurement error, not a discovery. Sitting out is a position, and often the correct one.

3. Your unrealized profit is probably fake.

On a thin book, the bid you're marking against isn't there when you go to sell. That take-profit that looks great on paper quietly bleeds when you actually try to fill. Mark at the price you can actually fill, not the price that makes your PnL screenshot look good.

The through-line: markets are mostly efficient, your backtest is overfit, and your paper gains are phantom. Respect all three and you stop bleeding.

The money comes from a narrow, proven edge repeated with discipline. Picks alone won't get you there.

[Dmitry Belaventsev @dizpers]

@zacodil Hmm didn't get it

[Dmitry Belaventsev @dizpers]

@Joi2James https://t.co/DmDVDyarZ5 ai text rpg -- browser / telegram

[Dmitry Belaventsev @dizpers]

17 years of software engineering experience.

Building a website for my psychotherapy practice. My first instinct was to code it from scratch. Second instinct: WordPress.

Ended up on a website builder platform. $10/month.

It already handles compliance rules I'd spend weeks researching, has a built-in CRM and transactional emails, appointment scheduling out of the box. All the stuff that has nothing to do with code and everything to do with running a practice.

Then I pointed Claude at the web editor through the browser and let it draft the content.

A senior engineer using a no-code platform and an AI writing the copy. Younger me would be embarrassed. Current me has a live website and zero headache.

Sometimes the most senior engineering decision is choosing not to engineer.

[Dmitry Belaventsev @dizpers]

why do engineering interviews take 2 to 4 months when you could just pay someone to work with you for a week and know more than 6 rounds of interviews would tell you.

seriously. hire them as a paid intern. give them tasks in the codebase, sit them next to the team. 5 days of shipping code together reveals more than any whiteboard session ever will.

week goes well? extend to 2 more weeks, bump the pay. still good? here's 4 more weeks at full rate. you just hired someone based on evidence instead of interview performance.

both sides get to evaluate. nobody wastes months on a process that tells you almost nothing about day to day work.

17 years in this industry. I've done technical interviews on both sides of the table, mentored engineers through job searches. the pattern is always the same: interviews measure interview skills. the job requires engineering skills. these overlap way less than people think.

thoughts?