# Cult of Coders > Cult of Coders is a software development agency that builds custom web apps, mobile apps, and AI-accelerated MVPs for startups and businesses. Delivery is led by senior engineers who use coding agents as part of the toolkit, not as a replacement for review and judgment. 100+ products delivered, 11+ years in business, $5M+ in client revenue generated. The team takes on a limited number of new engagements at a time so every project gets a senior engineer's full attention rather than a bench of juniors. The company is explicit about how AI fits into delivery: agents draft and scaffold, a senior engineer reviews and owns every line that reaches a client's repository, and client code is never used as training data. ## Services - [How we build with AI](https://www.cultofcoders.com/how): The AI-accelerated delivery process behind MVPs shipped in about 4 weeks for under $10K — rapid documentation, data modeling, coding, and a SaaS boilerplate that skips the parts every project repeats. - [All services](https://www.cultofcoders.com/services): The four services in one place. - [Web development](https://www.cultofcoders.com/services/web-development): Front-end and back-end systems on modern stacks, planned, built, and delivered end to end, plus the infrastructure they run on. - [Mobile development](https://www.cultofcoders.com/services/mobile-development): Native-quality iOS and Android apps built with React Native, one codebase for both stores, with the API behind them. - [Website design](https://www.cultofcoders.com/services/website-design): Custom-designed websites with real character, not templates. Designed and built by the same team. - [Staff augmentation](https://www.cultofcoders.com/services/staff-augmentation): Locally managed senior developers who plug into an existing team for extra capacity or a specialist opinion. ## Company - [About](https://www.cultofcoders.com/about): Who the team is and how the company works. - [Careers](https://www.cultofcoders.com/careers): Open roles and the application form. 100% remote, open to applicants across the EU. The page is the source of truth for which roles are live. - [Contact](https://www.cultofcoders.com/contact): Book a free strategy call. New engagements are limited to keep quality high. ## Work - [Portfolio](https://www.cultofcoders.com/portfolio): Case studies across industries, including startups that reached 1M users and enterprises that cut costs by 60%. ## Writing - [Blog](https://www.cultofcoders.com/blog): Every article, newest first. Long-form pieces on software delivery, design process, and AI-assisted engineering. - [Theodor Diaconu](https://www.cultofcoders.com/authors/theodor-diaconu): The byline behind the writing, and the profile the articles attribute to. ### Agents and AI-assisted engineering - [Fleets Need a Grader](https://www.cultofcoders.com/blog/article/fleets-need-a-grader): Six months of 2026 receipts on multi-agent swarms, and the one condition separating fleets that win from fleets that burn the budget. - [Coding Agents Hit Production. Production Hit Back.](https://www.cultofcoders.com/blog/article/coding-agents-in-production): The July 2026 field report — productivity numbers, the review crunch, the workflows that stuck, and the year's security incidents. - [Nobody Reads the Spec](https://www.cultofcoders.com/blog/article/nobody-reads-the-spec): Why more specification made agents worse, and the handful of spec-driven practices that survived contact with production. - [Your Agents Aren't Tired. You Are.](https://www.cultofcoders.com/blog/article/your-agents-arent-tired-you-are): Agent fatigue documented — the parallel-agent ceiling, review exhaustion, and the attention-budget playbook. - [Agentic Engineering in 2026](https://www.cultofcoders.com/blog/article/agentic-engineering-2026): A year of practitioner opinion distilled — context engineering, RAG vs. fine-tuning, multi-agent architecture, and why judgment is still the job. - [Your Code Got Cheap. Your Signature Didn't.](https://www.cultofcoders.com/blog/article/your-code-got-cheap): Open source ran on proof of work and AI made work free. What broke, and why the market started pricing the human who signs. - [Somebody Has to Run the Exploit](https://www.cultofcoders.com/blog/article/somebody-has-to-run-the-exploit): AI fixed 1,072 Chrome bugs and fabricated 54 of 55 advisories in one week. Why the reproducer is the invoice. ### Model economics and cost models - [The Harness Tax](https://www.cultofcoders.com/blog/article/the-harness-tax): Three 2026 measurements agree that the harness, not the model, decides the agent bill. Where the tokens go and why the cache ledger rules. - [A Billion Tokens for $6.36](https://www.cultofcoders.com/blog/article/deepseek-price-increase): What a real billion-token invoice decodes to, why the cache tier carries the exposure, and which model still has an exit. - [Same Score, 1/36th the Bill](https://www.cultofcoders.com/blog/article/deepseek-flash-same-score-cheaper-bill): DeepSeek V4-Flash-0731 checked against primary sources, including the 25.8-point jump nobody actually claimed. - [DeepSeek Shipped a Post-Train, Not a Model](https://www.cultofcoders.com/blog/article/deepseek-v4-flash-0731): The frozen-weight benchmark jump, the verbosity tax that eats the discount, and the 2x peak surcharge nobody priced in. - [The Day AI Pricing Fell Off a Cliff](https://www.cultofcoders.com/blog/article/gpt-5-6-luna-price-cut): The July 2026 GPT-5.6 price cut decoded — nano-parity fine print, cached-token math, who should switch and who should stay. - [The Real Cost of a $500 Fine-Tune](https://www.cultofcoders.com/blog/article/real-cost-of-500-fine-tune): A practical fine-tuning cost model covering data, verifiers, serving, maintenance, and five approval gates. - [What It Actually Costs to Self-Host Kimi K3](https://www.cultofcoders.com/blog/article/kimi-k3-self-hosting-engineering-gates): The 8-GPU hardware floor, provider speeds, rented clusters, owned servers, and the API break-even point. - [The Number Nobody Checked](https://www.cultofcoders.com/blog/article/the-number-nobody-checked): The claim that 9 in 10 startups fail, traced to an uncited PDF and a Bloomberg study that does not exist. ### Delivery and design - [How we use the design thinking process](https://www.cultofcoders.com/blog/article/design-thinking-process): Double-diamond design thinking end to end, and the discovery, research, design and validation methods behind it. - [How to fail an IT project](https://www.cultofcoders.com/blog/article/how-to-fail-an-IT-project): The five mistakes that sink IT projects, and the communication gaps tying them together. - [Increase sales with customer-centric design](https://www.cultofcoders.com/blog/article/customer-centric-websites): Why organization-centric design costs sales, with a Wijntransport case study. - [Automating healthcare websites](https://www.cultofcoders.com/blog/article/automating-healthcare-websites): Why healthcare sites automate, and what we learned building patient-facing tools under COVID-19 pressure. ## Open source - [BlueLibs](https://www.bluelibs.com/): The company's open-source framework. Stands up a production-ready project in minutes with data models, CRUD, a GraphQL API, live data, authentication, and uploads already wired. - [Runner](https://runner.bluelibs.com): A TypeScript-first toolkit of tasks, resources, events, and hooks, with retries, circuit breakers, and durable workflows built in. ## Optional - [Privacy Policy](https://www.cultofcoders.com/privacy-policy) - [Terms and Conditions](https://www.cultofcoders.com/terms-and-conditions)