Weekly Log

May 24, 2026

Backend

• Faced issues with OpenRouter billing; we need to either implement our own billing or find a workable solution for OpenRouter billing.

General

• Company legal structure chosen.
• Company registered.

May 17, 2026

Frontend

• Working MVP with auth, employee management, interaction logging, credits system, and now full EN/RU localization.

General

• Research on markets for business roll-ups, Company incorporation.

May 8, 2026

Frontend

• Implemented a billing mechanism.
• Moved from internal credit accounting to a model where the real balance lives on the OpenRouter side.
• Added support for converting the balance to Russian rubles.

Backend

• Hardened infrastructure security.
• Removed/hidden sensitive information about the slave machine used by the exec bot.
• Business impact: lower risk of leaking internal architecture and environment data; lower likelihood of abuse and incidents.
• Added BYOK support in per_user mode (with encryption).
• User keys can now be applied at the individual user level.
• Business impact: stronger compliance posture (enterprise requirements), higher trust from large customers, and better competitiveness for selling into regulated segments.

General

• Defined the company structure to support day-to-day operations and future investment goals.
• Analyzed the market for building a roll-up company in Russia across four segments, including accounting and real estate.
• Took the first steps toward a deal with an enterprise company: their needs are not fully formed yet, but they requested two more AI bots, one of them on OpenClaw.
• First failed attempt to roll out a large AI tool in another company due to a lack of authority for a strict rollout; we continue working on another product that may be more in demand.
• Implemented the first version of a bot for development management and engineering, based on OpenClaw + nanobot.

April 30, 2026

Backend

• This week, the focus was on security and isolation for running processes/services.
• Expanded the database schema: added new fields to link assistants with the orchestrator, plus an error status.

Frontend

• Added orchestrator integration for creating and managing Telegram bots.
• Fixed a status-check bug: the UI now correctly reads ActiveState from the orchestrator response.

General

• Submitted an application to YC Summer 2026.
• Moved the release date to June 27.

April 24, 2026

Backend

• Expanded LLM credential management to support both platform-provided and bring-your-own-key configurations, with tighter model compatibility controls.
• Added and improved core API operations for bot lifecycle management (invite rotation, router status, and router restart).
• Improved Telegram bot provisioning so per-user custom settings are stored and applied more reliably.
• Strengthened runtime stability through better network binding defaults and per-instance workspace behavior.
• Improved error reporting and diagnostics to make failures clearer and easier to troubleshoot.
• Refreshed technical documentation and API specs to align with the latest platform behavior.

Frontend

• Bot provisioning and teardown through an external orchestrator.
• Live status monitoring and one-click restart for bots.
• Shareable invite codes with copy and rotate functionality.
• Dynamic AI model selection from multiple providers (Anthropic, OpenAI, Google).
• Error tracking and visibility when bot setup fails.
• Interaction logs with token usage and cost breakdown.

General

• Pushing the launch back by 2–3 weeks to complete a solid technical solution, assemble a go-to-market strategy, and start feeding traffic to the current waitlists.

April 17, 2026

Backend

• New decentralized architecture implemented.
• Stateless Master, stateless Slave, and stateless Messenger router.
• Worst-case scenario is now limited to a single bot failure (or partial workspace failure for a particular bot).
• Can hold up to 50,000 users.

Frontend

• New versions of DarwinClaw office visuals implemented.
• Will post visuals next week.

General

• Reddit account boosted to start content marketing soon.
• Pikabu account reaches 100k+ views per post now.
• Product placement strategy defined.

Investments

• Investment requirements clarified.
• Money needed to run for the next year to reach breakeven in ~1 year.

Next

• Backend testing.
• Finish development of the frontend.
• Start contacting funds and business angels.

April 10, 2026

Backend

• The one-bot, multiple-instances backend architecture was rejected due to a single point of failure.
• It relied on a single router with user-to-instance mapping.
• We will try another approach: one instance with many workspaces.
• This should remove the single point of failure and significantly reduce RAM consumption.
• We already have experience running a single bot for 5,000 employees, so this approach should work well.

General

• Reddit account recovery has started.
• Applied to a startup accelerator.
• Joined the board for AI-ization of a $10M+ company.

April 8, 2026

Joined the board of a large company ($10M+ revenue) for AI-ization. First, complete AI-ization of all development. We will elevate the company to Level 1 in 2 months, then to Level 2. As always, there are technical difficulties, but the biggest challenge is mental: it is hard to break people's habit of coding :)

• Level 0 — ChatGPT is used instead of Stack Overflow.
• Level 1 — AI IDE (like Cursor) is used.
• Level 2 — Cloud AI pipeline (Design + ReAct agent pattern + QA).
• Level 3 — Super Product Owner (one product owner leads 3–5 products without Dev/QA teams).

April 8, 2026

You can try to do the same thing that NASA Artemis II ship does yourself here: https://darwinclaw.ai/MoonFlyBy/

April 6, 2026

Try NASA Artemis Lunar Flyby yourself. Can you make it?

April 6, 2026

Not related to DarwinClaw, but a very important event in orbit around the Moon should happen soon.

April 3, 2026

Minimalistic front

Week: March 30–April 5, 2026

Backend

• Master: node-less HTTP endpoints, /healthz, configurable port and auth.
• Slave: RAM allocation wired through slave, master logic, and machines template.
• Docs/READMEs and curl examples updated.
• Fully stateless version of DarwinClaw Back assembled (Master+Slave machines code).
• Next: implement one bot to many agents routing for B2B/Enterprise cases.

Frontend

• Minimalistic version of Front drafted. Screenshots will be posted here later.

General

• DarwinClaw.AI Scale invented to drive marketing.

April 1, 2026

Classifying Future Organizations by Efficiency

There's a famous classification of civilizations according to their level of development by Kardashev. The principle is simple: the more energy a civilization uses, the more advanced it is.

• 👉 Level 1 consumes all the energy of the planet + the energy of the star on the planet ~10^16 W
• 👉 Level 2 consumes all the energy of its star (Dyson sphere) ~10^26 W
• 👉 Level 3 consumes all the energy of the galaxy ~10^36 W

It's clear already that the lion's share of future energy will be consumed by AI.

But how do we calculate AInization (level of AI integration at all levels of the company) in a specific organization?

Such an indicator already exists, essentially it's the consumption of tokens, or to be more precise, we want to understand how much the organization has transitioned to AI in terms of maximum automation of all processes and replacing all computer operators (any specialty, from a lawyer to a programmer) with AI wherever possible.

Then the metric will be money, or rather % of expenditures on AI (tokens or your own neural networks) out of total payroll.

I observe many organizations (which have already implemented and continue to implement AI at various levels) and I think that this scale would suit (by analogy with Kardashev):

• 👉 Level 0 ~1/4 of total payroll (the organization has begun automation)
• 👉 Level 1 ~1x total payroll (the organization is partially automated)
• 👉 Level 2 ~4x total payroll (minimal staff for AI management)
• 👉 Level 3 ~10x total payroll (the organization is almost staffless)

Just like in Kardashev's system (where we only know of civilizations below Level 1), I don't think there are any organizations above Level 0 yet, but they will definitely appear in the coming years 😎 Let's call this the AInization scale by DarwinClaw 🤖

Week: March 23–29, 2026

Backend

• Built and stabilized a Master + Slave environment for managing nanobot instances on remote nodes.
• Implemented RAM-based capacity scheduling in the orchestrator (ramCapacityMB) with automatic node selection and no capacity available protection on overflow.
• Moved instance creation to a globally unique username-uuid format.
• Updated slave-side instance validation and port assignment logic; improved systemd service templates.
• Updated machine configuration files (machines.json and template) and expanded deployment/operations docs (README_SLAVE_DEPLOY.md, curl examples, firewall checks).
• Ran smoke checks for create/list/status/restart/delete, plus network and capacity diagnostics.

Frontend

• Released a draft version of the client personal office.
• Released a draft version of the client disk feature.

General

• Refined the version roadmap.
• Migrated all sites and landings to DarwinClaw Moon Alpha management.
• Discussed DarwinClaw rollout for two companies.