Building intelligent machines that see and communicate. Passionate about AI, machine vision, and software engineering. Entrepreneur and innovator. Let's connect

Izmir
Joined December 2010
Follow if you want a front row seat to AI, machine vision, Web3, and the craft of building elegant machines that print value. Curiosity in, compounding out. #AI #Blockchain #Startups
1
8
Writing hack: Specific > general. Vague: “We improved the app.” Specific: “Cold-start time fell from 2.4s → 1.1s, crash rate −38%.” Numbers, verbs, before/after. Your readers don’t need adjectives, they need evidence.
1
Ingenuity tip: Use Fermi estimates to get fast answers. Break a big unknown into rough pieces, round aggressively, multiply. You’ll often be within 10×, good enough to decide. Example: “How many pianos in NYC?” Population × households × piano prevalence.
2
The next unicorns combine three loops, data generation, agentic execution, and human verification. Closed‑loop learning prints moats. #Startups #AI
Isomorphic Labs raised $600M to push AI‑designed drugs toward trials. The bio‑software stack is getting funded like software. #BioAI #Startups #VC
1
1
Startup moat in 2025, not prompts, not wrappers. It is data network effects, policy engines, evals, distribution, and the right infra contracts. #Startups #AI #Strategy
Gartner says 40 percent of “agentic AI” projects get scrapped by 2027. Translation, build agents where the data, tools, and ROI are crisp, then scale. #AI #Agents #Product
1M businesses using OpenAI, with millions of seats and agents graduating from pilots to production. The AI adoption S‑curve just found its steep part. #AI #SaaS #Enterprise
AWS x OpenAI, multi‑year, big‑number compute pact. Read that as scale, availability, and faster iteration for frontier models. Compute is the new go‑to‑market. #AI #Cloud #Infra
Suleyman Kivanc EKICI retweeted
I wish my parents had taught me this. But my mom refused to talk about money, because it's all her parents cared about. I understand her intentions, but it was a dramatic overcorrection.
2
10
2
219
Elegance is a ruthless kindness to future you. One concept, many uses. #Engineering #Design
Prompt library idea: keep only prompts that survived production. Everything else is poetry. #LLM
Synthetic data is creatine, not dinner. Great for gains, not a substitute for real edge cases. Cycle responsibly. #AI #Data
1
Hiring signal: T shaped builders with one deep obsession, a second domain they respect, and business taste. Pair a roboticist with a salesperson, watch magic. #Startups #Hiring
Sales for AI: do not sell mean improvement, sell reduced variance. CFOs respect variance control. Screenshots later, risk bands first. #Sales #AI #B2B
Metric to obsess over this quarter: Time to First Magic Moment. Everything else is commentary. What is your TTFMM today? #Growth #Product #Startups
NFTs are machine readable warranties. Scan, validate, transfer, expire, redeem. Less monkey, more maintenance. What real world asset would you tokenize first? #Web3 #NFTs #RWA
4
Outcome pricing beats seat pricing for AI. Meter by risk, settle by results, upsell by autonomy. Tier names matter more than you think: Explore, Control, Commit. Steal this. #SaaS #AI #Pricing
Ship an eval scoreboard as a first class feature. Every merge triggers tests, a changelog, and a delta on business metrics. Pin it, brag responsibly, iterate. Want my YAML starter pack? Comment evals. #MLOps #LLM #SaaS
If your product only accepts emails, you built for humans. If it also accepts keys, you built for humans with bot armies. Design for agents that pay, rate limit, and get slashed. Who is shipping this already? #AI #Blockchain #Agents
1
Your real moat in 2025 is distribution with taste. A community that trusts you to filter noise becomes an engine for compounding truth and revenue. Curate viciously. Do you agree? #Startups #Brand #Community
1