Stratagem
What Is This About?
if you're reading this, you're watching me build byzantex in real-time. might fail spectacularly. might build something massive. either way, this is the unfiltered version.
the mission: ai-powered verification tools for semiconductor companies. specifically, we're solving the regression triage bottleneck that wastes 30+ minutes daily for verification engineers.
the approach: look, i hate building for VCs. but this time i'm playing their game to understand what it actually takes. applying to accel atoms, antler, ef, and spc - not just for the money, but to learn how billion-dollar companies actually get built.
the reality check: before writing ANY code, i spent 10 weeks talking to 70+ verification engineers. ARM, google, qualcomm, sifive, indian startups - the whole spectrum. found out that even google and qualcomm built internal scripts for this problem. which means no good commercial solution exists.
that's our opening.
Pincer Movement: Why We're Attacking From Both Sides
learned about two strategies for selling enterprise software - top-down (sell to executives) or bottom-up (sell to users). most people pick one.
we're doing both. here's how:
bottom flank - the engineers:
- talk to verification engineers directly
- understand their actual pain (not what i think it is)
- build poc based on real workflows
- get them to champion the tool internally
- create demand from the ground up
top flank - the decision makers:
- armed with engineer validation, go to CTOs/VPs
- show them: "your team already wants this, here's the ROI"
- it's not a cold pitch anymore - it's engineer-recommended
- close design partnerships and contracts
why this works: engineers validate the problem. CTOs sign the checks. by approaching both, we surround the sale.
where we're at:
- talked to 70+ engineers, had 20+ deep conversations
- got validation from ARM, google, qualcomm, sifive, mindgrove
- built a working poc (tested on opentitan logs)
- having active discussions with mindgrove and incore right now
The 3 Classes: How We're Going To Scale This Thing
breaking the next 18 months into three phases. each phase targets different customers.
Phase 1: Design Partners (Months 1-6)
who we're going after: indian risc-v startups - mindgrove, incore, morphing machines, l&t semiconductors
why these companies:
- small teams (5-20 verification engineers)
- no internal tools teams (can't build this themselves)
- high pain, fast decisions
- they move FAST (no corporate bureaucracy)
the deal: offer lifetime design partner benefits - early access, co-develop features, and preferred pricing forever. Basically, be our first customers and let us build trust in the sector.
What we're doing:
- building poc specifically for their workflows
- getting testimonials and case studies
- using this as social proof for funding and phase 2
The goal:
- sign 2-3 design partners
- validate product-market fit
- hit $50k arr
- raise pre-seed ($500k-750k)
current status (Dec 24, 2025):
- POC is built. tested on opentitan logs (a Google-backed risc-v project)
- demo video recorded
- scheduling poc demos with mindgrove and incore in january
- target: first design partner signed by february
Why phase 1 matters: This proves we can actually solve the problem. not just talk about it. builds credibility with U.S. startups later. creates pricing reference. generates testimonials for when we go after bigger companies.
Phase 2: Early Adopters (Months 6-12)
Who we're targeting:
- india: tessolve, mirafra, moschips
- usa: sifive, akeana, rivos, ampere, tenstorrent, ventana
- global: alphawave, other 50-200 person teams
Why these companies:
- bigger teams (20-50 verification engineers)
- have some internal scripts but they're brittle and unmaintained
- have budget for tools ($50k-150k/year)
- would actually pay for a polished, maintained solution
The approach: Lead with design partner proof. "Mindgrove already uses this." offer premium pricing ($3k-5k/month per team). Build case studies showing ROI. Expand features based on feedback. Add integrations (Jenkins, Jira, Slack).
The goal:
- get 10-15 paying customers
- hit $300k-500k arr
- prove the gtm motion works
- raise seed round ($2-3m)
Why phase 2 matters: validates that people will actually PAY. scales beyond just design partners. proves the product works for multiple companies, not just one. builds revenue for fundraising.
Phase 3: Scale + Enterprise (Year 2-3)
Who we're going after:
- 50-100 engineer verification teams
- design service companies
- adjacent markets: arm divisions, automotive (renesas), aerospace
- eventually: nvidia, Qualcomm, and amd (but year 3-5, not now)
Why these companies:
- large verification organizations
- existing tools/processes (harder to displace)
- enterprise budgets ($200k-500k+ annually)
- need: security certs, enterprise slas, custom integrations
The approach: Lead with customer proof (15+ companies already using us). enterprise sales motion (6-12 month cycles). Add enterprise features (SSO, audit logs, on-prem). Build partnerships with EDA vendors. expand to adjacent workflows beyond just regression triage.
The goal:
- sign 50+ teams total
- hit $2-4m arr (year 2)
- path to $10m+ arr (year 3)
- raise series a ($8-12m)
NOT targeting yet: large enterprises like Nvidia, Qualcomm, Intel, and Samsung. Why? 18+ month sales cycles. mature internal tools. high switching costs. We'll go after them in years 3-5, after we've proven ourselves.
Why phase 3 matters: This is venture-scale. $10m+. proves defensibility—that larger companies will choose us over building internally. creates network effects. makes us the industry standard. enables Series A+ fundraising.
Current Status (Dec 24, 2025)
What's done:
reached out to 70+ verification engineers
had 20+ in-depth conversations
validated the problem: regression triage wastes 30+ minutes daily
key insight: even google/qualcomm built internal scripts (no good commercial solution)
sized the market: tam $2-3b, sam $800m-1.2b, som $4-5m arr by year 3
pitch deck done (honestly pretty solid)
built the poc: ai-powered regression triage tool
tested on opentitan verification logs
categorized 127 tests, 23 failures
reduced triage time: 30 min → under 2 min (93% reduction)
recorded demo video
What's next (next 30 days):
- jan 1: submit applications (accel atoms, antler, ef, spc bangalore)
- jan 2-7: poc demos with mindgrove, incore
- jan 15: hopefully sign first design partner
- jan 31: accelerator interviews (fingers crossed)
Q1 2026 Goals:
- feb: iterate poc based on design partner feedback
- march: ship mvp to first customers
- march 31: hit $50k arr
- q1 end: close pre-seed round ($500k-750k)
What I'm Learning Along The Way
1. Validation Beats Ideas Every Time
spent 2 months talking to 70+ engineers before building anything. Most founders do it backwards—build first, validate later. Those customer conversations revealed the REAL problem, not what I assumed it was.
2. Speed Is A Competitive Advantage
went from zero semiconductor knowledge to 70+ conversations to working POC in 10 weeks. during the holidays while everyone else was partying. Execution speed = founder credibility.
3. Bottom-Up Actually Works
Engineers WANT better tools. They'll champion you internally if you solve their pain. way easier than cold calling CTOs.
4. Open-Source Is A Goldmine
OpenTitan gave us real verification logs without any NDA. could build POC without waiting for design partners to share data. plus credibility: "tested on the Google-backed OpenTitan project."
5. Small Companies Move Fast
Indian startups (Mindgrove, Incore) respond in days. Giants (ARM, Qualcomm) take weeks or months. That's why phase 1 focuses on the fast movers.
The Honest Truth About What Could Go Wrong
Risks I'm watching:
- design partners might not commit (30% chance)
- poc might not resonate with users (20% chance)
- accelerators might reject us (40% chance, honestly)
- market might be smaller than expected (15% chance)
- might burn out before hitting pmf (25% chance)
Why I might actually succeed:
- deep customer validation (70+ conversations is more than most)
- clear, quantified problem ($125k/year wasted per team)
- working poc (not just slides and promises)
- fast execution (10 weeks, during holidays)
- right market timing (india semiconductor boom, risc-v adoption)
- underserved segment (small companies can't build internal tools)
the commitment: next 6 months, 100% focused on this. no side projects. no distractions. Either build something real or fail publicly. documenting everything so others can learn from it.
Follow Along
updating this blog weekly with:
- milestones hit (and missed)
- customer conversations and what i'm learning
- accelerator application outcomes
- design partner updates
- revenue numbers
- fundraising progress
- honest failures and pivots when they happen
No bullshit. No fake wins. just the real journey.
Current Asks
If you're a verification engineer: want to see a demo? → mihir@byzantex.online
want to be a design partner? → let's talk
If you're at an accelerator: reviewing our application? → this blog is the full context.
want to dig deeper? → happy to chat
If you're an investor: interested in pre-seed? → we're raising Q1 2026.
want early access? → let's connect
If you're building something: learn from this journey.
customer discovery is greater than everything else.
ship fast, iterate faster
Links
- website: byzantex.online
- email: mihir@byzantex.online
- phone: +91 9920014592
last updated: December 24, 2025
next update: January 7, 2026 (after application submissions)
Let's build something real.
— mihir
Would you like me to create a punchy LinkedIn post to announce this new blog entry?