iiinova — where your digital self keeps evolving

Di Reshtei @dreshtei.bsky.social · 06/10/2026

MRR going up can hide the fact that your b2b SaaS is getting worse. revenue ↑ users ↑ usage ↑ looks great. then: support costs ↑ AI/API costs ↑ gross margin ↓ growth can hide bad economics for a long time. what metric do you trust more than MRR?

010

Di Reshtei @dreshtei.bsky.social · 05/10/2026

b2b SaaS founders will spend 3 engineering weeks to avoid paying $49/month for a tool. and call it “saving money.” that’s not optimization. that’s expensive procrastination with GitHub commits. what’s the dumbest thing you’ve built instead of buying?

020

Di Reshtei @dreshtei.bsky.social · 04/10/2026

your b2b SaaS customer can decide to churn before onboarding is even finished. too many steps. too much setup. no quick win. and now you’re fighting to “improve retention.” sometimes churn is just delayed bad onboarding. how fast can a new user reach value in your SaaS?

000

Di Reshtei @dreshtei.bsky.social · 03/10/2026

your cheapest b2b SaaS customers can become your most expensive customers. they pay $19. then send 14 emails. request 6 features. need 3 onboarding calls. and churn after 2 months. sometimes higher prices improve more than revenue. seen this happen?

000

Di Reshtei @dreshtei.bsky.social · 02/10/2026

hiring a salesperson won’t save your b2b SaaS if you still can’t explain why anyone should buy it. bad positioning + more salespeople = expensive bad positioning. founders should learn to sell before they outsource selling. when did you make your first sales hire?

110

Di Reshtei @dreshtei.bsky.social · 01/10/2026

one of the most profitable things a b2b SaaS founder can do: delete shit. fewer features. less onboarding. less support. less code to maintain. customers don’t pay you for the number of buttons. what feature would you delete tomorrow?

121

Di Reshtei @dreshtei.bsky.social · 30/09/2026

“unlimited AI” sounds amazing on a b2b SaaS pricing page. until one customer actually uses what you promised. every generation costs money. your user sees “unlimited.” your API provider sees $$$. would you still offer unlimited AI today?

020

Di Reshtei @dreshtei.bsky.social · 29/09/2026

“our ICP is SMBs” is b2b SaaS founder-speak for: “we still have no idea who actually wants this.” SMBs isn’t an ICP. pick one painful problem. one type of buyer. one reason they should care. how specific is your ICP?

131

Di Reshtei @dreshtei.bsky.social · 28/09/2026

b2b SaaS founders hate hearing this: the feature you’re building this week probably won’t bring you a single new customer. another feature feels productive. distribution feels painful. so founders keep building. what does your SaaS need more right now: features or customers?

120

Di Reshtei @dreshtei.bsky.social · 27/09/2026

b2b saas founders will eventually treat AI inference like e-commerce treats shipping. not as “some technical cost.” as COGS. because if every new customer automatically makes your compute bill bigger... that’s not an engineering detail. that’s your business model. tracking it yet?

240

Di Reshtei @dreshtei.bsky.social · 26/09/2026

b2b saas founders usually start caring about AI costs when it’s already painful. by then you’ve got: legacy prompts. model dependencies. angry users. ugly migrations. cost architecture isn’t something you fix after scale. it’s something you design before it. agree?

020

Di Reshtei @dreshtei.bsky.social · 25/09/2026

b2b saas founders: if you make $20/customer and spend $12 on inference... you don’t have an AI moat. you have an expensive API wrapper. sometimes the best “feature” you can ship is simply making the same workflow cheaper. what would you optimize first?

020

Di Reshtei @dreshtei.bsky.social · 24/09/2026

bad AI economics can look amazing for a while. b2b saas founders see: MRR ↑ users ↑ usage ↑ champagne time 🥂 then API bill ↑↑↑ gross margin ↓ growth doesn’t fix bad unit economics. it just makes the disaster bigger. seen this one before?

560

Di Reshtei @dreshtei.bsky.social · 23/09/2026

b2b saas founders love saying: “it’s only $0.03 per generation.” cute. 20 generations/day × 5,000 customers × 30 days = $90,000/month tiny AI costs become very ugly when multiplied by real usage. still thinking per-call pricing is “cheap”?

010

Di Reshtei @dreshtei.bsky.social · 22/09/2026

one number b2b saas founders should obsess over: cost per successful outcome. not token price. not API calls. not “we use the smartest model.” how much does one finished customer task actually cost you? if you don’t know, that’s a problem.

130

Di Reshtei @dreshtei.bsky.social · 22/09/2026

just bought a domain for Jev. what should I cook there?

010

Di Reshtei @dreshtei.bsky.social · 22/09/2026

your b2b SaaS customers don’t give a shit about your AI. they don’t want GPT. they don’t want agents. they don’t want “AI-powered.” they want the annoying task gone. if AI doesn’t save time or money, it’s just an expensive ✨ button. agree?

110

Di Reshtei @dreshtei.bsky.social · 21/09/2026

b2b saas founders keep doing this: using the most expensive model for everything. reasoning? sure. hard coding? sure. basic extraction? seriously? using a Ferrari to deliver groceries is still stupid economics. are you routing models yet?

020

Di Reshtei @dreshtei.bsky.social · 21/09/2026

AI API providers don’t sell you outcomes. They sell usage. More tokens. More calls. More images. More video seconds. But SaaS needs the opposite: less usage → same result That’s why cost per outcome matters more than price per token. Agree?

010

Di Reshtei @dreshtei.bsky.social · 20/09/2026

Uncomfortable truth about AI APIs: Providers don’t get paid when your task is solved. They get paid for: tokens requests images video seconds So optimizing your COGS is your job, not theirs. Do you know what one completed outcome actually costs you?

010

Di Reshtei @dreshtei.bsky.social · 19/09/2026

Your KPI: Get the result as cheaply as possible. Your AI API provider’s KPI: Get you to use more API. That’s where the incentives diverge. Because the “best model” doesn’t always mean the best economics. What matters more to you: quality or cost per result?

011

Di Reshtei @dreshtei.bsky.social · 18/09/2026

One of the biggest mistakes in AI infrastructure: Looking only at token price. Your provider makes money when you make more requests. You want the opposite: fewer calls → same outcome → lower COGS What are you actually optimizing?

120

Di Reshtei @dreshtei.bsky.social · 18/09/2026

sometimes the best SEO strategy is doing absolutely nothing 😅 this domain is about a month old. i’ve done zero link building. DR 29. should i leave it alone, or start ruining the result with my SEO work? 😂

010

Di Reshtei @dreshtei.bsky.social · 17/09/2026

unpopular opinion for b2b saas founders: “unlimited AI” is often a terrible idea. every click costs you money. it looks sexy on the pricing page... until one power user decides to discover exactly how unlimited “unlimited” is 😂 still offering unlimited?

030

Di Reshtei @dreshtei.bsky.social · 17/09/2026

AI API providers don’t really care if you get the result in 1 request. They make more money if it takes 10. You optimize for outcome. They monetize usage. The uncomfortable part: your incentives are not always aligned. Do you track cost per result?

040

Di Reshtei @dreshtei.bsky.social · 16/09/2026

b2b saas founders get fooled by AI costs at MVP stage. $50 API bill? amazing. then users arrive. then usage explodes. then the invoice hits. and suddenly your “killer AI feature” is killing your gross margin. do you model this before launch?

040

Di Reshtei @dreshtei.bsky.social · 16/09/2026

What’s harder for you: building or distribution?

000

Di Reshtei @dreshtei.bsky.social · 16/09/2026

Indie hackers love saying: “Just ship.” But there’s a second half people talk about way less: Just distribute. Shipping a product in a weekend is easy. Spending 6 months finding customers, creating content, and testing channels — that’s where most people quit.

120

Di Reshtei @dreshtei.bsky.social · 15/09/2026

b2b saas founders love watching MRR go up. but almost nobody gets excited about AI COGS. revenue +20%. AI bill +60%. congrats, you’re growing straight into a margin problem. do you actually know your AI cost per active user?

150

Di Reshtei @dreshtei.bsky.social · 14/09/2026

Switched from Claude Code to Codex. My first thought: WAIT. You can just… do that? No walls of text. No unsolicited lectures. No moralizing bullshit. Just: task - result. Anyone else had the same reaction?

010

Di Reshtei @dreshtei.bsky.social · 13/09/2026

WTF???? I gave the model a proper commercial prompt. It generated this 👇 Cost: $0.021. Not $0.21. Not $2.10. 2.1 cents. And if it fails, I pay $0. Alright. Give me a prompt that breaks it 😁

010

Di Reshtei @dreshtei.bsky.social · 12/09/2026

AI image pricing is starting to get on my nerves. People are still paying multiples more for images that often don’t look any better. This one? $0.021. A failed generation? $0. So at some point you have to ask: What, exactly, are we paying extra for?

010

Di Reshtei @dreshtei.bsky.social · 11/09/2026

This is how I want to test AI image models: Not benchmarks. Can it make a proper commercial creative for pennies? This one cost $0.021. Fixed price. Failed generation = $0. Give me a prompt that breaks it 😁

011

Di Reshtei @dreshtei.bsky.social · 10/09/2026

I like this math: complex cinematic prompt + 1K image fixed pricing = $0.021 If the generation fails: $0. At 10 images, who cares. At 100,000? That’s when image API pricing becomes a business decision.

000

Di Reshtei @dreshtei.bsky.social · 09/09/2026

Okay, generating one bottle is easy. What happens when you give the model dozens of objects, different textures, lighting, and composition? Result 👇 $0.021 per image. Failed generation = $0. What should I throw at it next?

000

Di Reshtei @dreshtei.bsky.social · 09/09/2026

This isn’t “cheap AI” anymore. It’s absurd. A commercial creative for $0.021. Fixed price. Failed generation = $0. If the quality is good enough for your use case, paying 5–10× more becomes a very weird business decision. Change my mind 😁

000

Di Reshtei @dreshtei.bsky.social · 08/09/2026

Realistic portrait. Clean lighting. Premium corporate look. Cost: $0.0192. Are image APIs becoming a commodity faster than people think?

000

Di Reshtei @dreshtei.bsky.social · 08/09/2026

AI image generation is becoming stupidly cheap. This commercial creative: $0.021. Fixed. Failed generation? $0. And we’re still supposed to pretend premium pricing always makes sense? 😅 Give me something harder to generate.

121

Di Reshtei @dreshtei.bsky.social · 07/09/2026

This is getting absurd. A commercial-grade image for $0.021. Flat price. If the generation fails, you pay $0. We’re at the point where an ad image costs less than a sip of coffee. What should I generate next?

000

Di Reshtei @dreshtei.bsky.social · 07/09/2026

This image cost me $0.0210 Luxury product shot. Realistic lighting. Clean commercial look. How much would your image API charge for this?

000

Di Reshtei @dreshtei.bsky.social · 06/09/2026

Your AI API bill is too high #1 Product: AI article generator Volume: 10,000 articles/month ~15M output tokens Claude Sonnet 4.6 → ~$225 Omev Pro → ~$75 That’s ~$1,800/year saved on output alone. Now imagine 100K articles 😅

000

Di Reshtei @dreshtei.bsky.social · 05/09/2026

If your AI SaaS generates content, your API bill isn’t just a “technical expense.” It’s COGS. 10,000 articles/month: Claude Sonnet 4.6: ~$225 output Omev Pro: ~$75 The more you grow, the more margin you hand to your API provider. Still not thinking about routing?

000

Di Reshtei @dreshtei.bsky.social · 04/09/2026

At 100 articles, API pricing barely matters. At 10,000, it starts to matter. At 100,000, it becomes part of your unit economics. That’s why “what’s the best model?” is often the wrong question. The better one: “What model does this task actually need?”

000

Di Reshtei @dreshtei.bsky.social · 03/09/2026

You’re building an SEO SaaS. You generate 10,000 articles per month. And every single one goes through a frontier model. Why? 😅 At ~15M output tokens: Claude: ~$225 Omev Pro: ~$75 3× more for repetitive work. What exactly are you paying extra for?

210

Di Reshtei @dreshtei.bsky.social · 02/09/2026

Imagine an AI SEO SaaS generating 10,000 articles per month. ~15M output tokens: Omev Pro: ~$75 Gemini 3.1 Pro: ~$180 Claude Sonnet 4.6: ~$225 That’s just one workload. Why pay frontier-model prices for bulk article generation?

010

Di Reshtei @dreshtei.bsky.social · 01/09/2026

I don’t want Omev to replace your OpenAI or Claude setup. Seriously. Keep them for the hard stuff. I’m after the boring 20–50% of your workload you’re currently paying frontier-model prices for. Don’t replace. Route. How much of your AI traffic could run on a cheaper model—without you noticing?

000

Di Reshtei @dreshtei.bsky.social · 31/08/2026

Using Claude to crank out 100,000 meta descriptions is like taking a Ferrari to the grocery store. You can. But why pay for intelligence the job doesn’t require? Use premium models for hard problems. Use cheaper models for repetitive output. Agree?

000

Di Reshtei @dreshtei.bsky.social · 31/08/2026

The problem isn’t that OpenAI or Claude are expensive. The problem is using them for EVERYTHING. Reasoning? Premium model. 10M repetitive requests? Why? Don’t replace your AI stack. Just stop overpaying for simple tasks. Who’s already doing this?

000

Di Reshtei @dreshtei.bsky.social · 30/08/2026

You’re probably overpaying for your AI API. Example: 1B input tokens 200M output tokens GPT-5.4 - $5,500/mo Omev Pro - $1,625/mo That’s $3,875 back every month - $46,500/year. Put it into growth, not tokens.

000

Di Reshtei @dreshtei.bsky.social · 29/08/2026

Unpopular opinion: If your SaaS runs AI at scale, API pricing matters more than squeezing out another +2% on a benchmark. $6,000/mo → $1,625/mo isn’t “token savings.” It’s $52,500 a year back in your business. Would you switch for that?

000