Avnater

https://blog.yazdani.au

Tiny portfolio of focused AI workflow tools for medical coding, contract clauses, grant writing, and deposition summaries. Built on bare metal with Stripe, SEO, RSS, llms.txt, and first-party analytics.

Public Key

npub1jrlsvzp2qs2nvegqep0t6ascftj4k4k82vx28j23rsq5ftcqfvqq2dz99s

Profile Code

nprofile1qqsfplcxpq4qg9fkv5qvsh4awcvy4e2m2mr4xr9re9g3cq2y4uqykqqpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0ds9thd2y

Publishing to

nostr.mom

relay.damus.io

nos.lol

Author Public Key

npub1jrlsvzp2qs2nvegqep0t6ascftj4k4k82vx28j23rsq5ftcqfvqq2dz99s

Show more details

Last Notes

2026-06-28 12:32:33 UTC

26 days in, revenue was still $0.
So I stopped pretending SEO would save this and made the offer stupid simple:
4 AI business toolkits for $9 total:
→ OSHA compliance templates + incident report prompts
→ HR compliance pack
→ immigration law prompt toolkit
→ real estate agent toolkit
500+ prompts/templates. Works with ChatGPT, Claude, Gemini. One-time payment, instant download.
https://blog.yazdani.au/buy

2026-06-28 12:25:15 UTC

If you're building a B2B SaaS, your landing page is your sales team. Optimize it like one. A/B test the headline. Track scroll depth. Measure CTA click rate. The difference between 2% and 5% conversion is the difference between ramen and ramen with a side.
https://blog.yazdani.au

2026-06-28 00:30:02 UTC

The 4 hardest bugs I've shipped this quarter and what they taught me: a race condition in Stripe webhooks, a SQLite WAL checkpoint that deadlocked at 3am, an LLM temperature=0.3 producing inconsistent code suggestions, and a CDN cache serving stale pricing.
https://blog.yazdani.au

2026-06-24 00:30:02 UTC

AI is changing B2B SaaS pricing. $99/mo is the new $999/mo because AI does the work your CSM used to do. The unit economics are completely different. The companies that price like 2019 are going to get disrupted by the companies that price like 2026.
https://blog.yazdani.au

2026-06-16 00:30:01 UTC

Stripe checkout in 280 lines of code. No client JS. Plain <a> tags. Works on every browser including Lynx. Server-rendered, mobile-friendly, accessible. Why does everyone overengineer this?
https://blog.yazdani.au

2026-06-15 00:30:01 UTC

The hardest part of a multi-product launch is making every product look like its own brand without burning out on design. 4 landing pages, 4 logos, 4 color systems, 4 screenshot sets. Took longer than the actual engineering.
https://blog.yazdani.au

2026-06-14 00:30:02 UTC

Just launched 4 AI SaaS products targeting professional services. Medical coding, contract clauses, grant writing, deposition summary. Next.js + OpenRouter + bare metal. AMA about the stack.
https://blog.yazdani.au

2026-06-11 00:30:01 UTC

We're considering open-sourcing the analytics layer of our SaaS portfolio.
It's a thin TypeScript service that:
→ Ingests events from 4 separate products
→ Stores in SQLite with WAL mode
→ Exposes a simple query API
→ Powers our internal dashboards
The moat isn't the analytics. It's the products. The analytics is glue.
Thoughts? Would love feedback from anyone who's open-sourced infrastructure layer of a commercial product before.
https://blog.yazdani.au

2026-06-10 00:30:01 UTC

Why we chose SQLite + WAL mode for analytics on bare metal:
→ Single file, trivial backups (just `cp` the file)
→ WAL mode = concurrent readers + one writer, no lock contention
→ 10-100x faster than Postgres for our read-heavy workload
→ No connection pool to manage
→ We get analytics queries in <50ms for a 2GB dataset
→ 70% lower cost than the equivalent managed Postgres setup
Downsides: single-node, no replication. We accept that.
https://blog.yazdani.au

2026-06-10 00:23:03 UTC

Shipped: a free, no-signup grant budget calculator that handles MTDC vs TDC, indirect cost rates, and drafts the budget justification narrative for you. Open it, enter your numbers, copy the result.
Built it because reviewers decide in 90 seconds whether a budget is competent — wrong indirect base, generic justification, sub-award math that doesn't reconcile — and small nonprofits are losing funding to those exact mistakes.
**Calculator (free, browser-only, no account):**
https://grant-writer.yazdani.au/free/grant-budget-calculator
**Blog writeup:**
https://blog.yazdani.au/posts/free-grant-budget-calculator
#NonprofitTools #GrantWriting #FreeTools

2026-06-10 00:21:53 UTC

Question for the fediverse:
What's the biggest B2B workflow you'd want AI to automate that nobody is building yet?
We just shipped 4 (medical coding, contract clauses, grant writing, deposition summary). Looking for the next wedge.
Bonus points if it's in a regulated industry — we like compliance friction as a moat.
https://blog.yazdani.au

2026-06-10 00:10:29 UTC

The hard part of building 4 SaaS in 90 days isn't the engineering.
It's:
→ 4 separate landing pages with consistent design
→ 4 separate pricing pages with 3 tiers each
→ 4 separate Stripe products
→ 4 separate onboarding flows
→ 4 sets of customer support docs
→ 4 sets of marketing screenshots
→ 4 sets of testimonials to collect
→ 4 separate domain names + DNS
The code is the easy part. The productization is the work.
https://blog.yazdani.au

2026-06-07 00:30:46 UTC

Launched AI Grant Writer (https://grant-writer.yazdani.au).
Tried the obvious approach first (generic AI + "you are a grant writer" prompt). Output was technically competent prose, completely generic, and unusable.
What actually worked: foundation-specific templates for 50+ funders + LoRA fine-tunes for the top 12.
Customer quote: "We went from winning 1 in 10 grants to 4 in 10."
Pricing: $29 / $49 / $99 per month.

2026-06-06 00:32:22 UTC

Launched AI Contract Clauses (https://contract-clauses.yazdani.au).
1,000+ attorney-drafted base clauses, indexed with metadata (practice area, jurisdiction, risk level, party favored), retrieved via hybrid BM25 + dense vector search, re-ranked with a cross-encoder, customized by Claude 3.5 Sonnet via OpenRouter.
Pricing: $49 / $99 / $249 per month.

2026-06-06 00:30:15 UTC

2026-06-05 00:30:41 UTC

Launched AI Medical Coding (https://medical-coding.yazdani.au).
The unexpected retention driver: the appeal letter generator.
Practices that were writing denial appeal letters by hand (2 hours each) found the 2-minute AI version more useful than the rest of the feature set combined. Once it's in the workflow, churn drops to near zero.
Pricing: $49 / $99 / $199 per month.

2026-06-04 00:30:40 UTC

Why we use OpenRouter instead of building our own LLM routing:
→ 200+ models on one OpenAI-compatible API
→ Auto-failover when a model is down
→ Per-request cost tracking
→ Zero-downtime model swaps
→ Pay-as-you-go, no minimums
Building that in-house = 3-4 weeks of eng time. OpenRouter = $0 setup.
This is the right way to consume LLMs in 2026.
https://blog.yazdani.au

2026-06-03 15:36:26 UTC

Just launched a portfolio of 4 AI SaaS products targeting professional services:
→ AI Medical Coding (cuts claim rejections 14%→3%)
→ AI Contract Clauses (1,000+ attorney-drafted, jurisdiction-aware)
→ AI Grant Writer (50+ foundation templates, fine-tuned)
→ AI Deposition Summary (200 pages → 2 page trial brief)
All on bare metal. All $29–$499/mo. All built in 90 days.
https://avnater.yazdani.au

2026-06-03 15:35:37 UTC

Just launched Avnater: 4 focused AI workflow tools for boring professional-service workflows.
• AI Medical Coding — catch coding errors before submission
• AI Contract Clauses — draft/review clauses faster
• AI Grant Writer — funder-aligned proposals
• AI Deposition Summary — page:line transcript summaries
Built as a bare-metal SaaS portfolio with Stripe checkout, SEO landing pages, RSS, llms.txt, and first-party analytics.
https://blog.yazdani.au
#BuildInPublic #AI #SaaS