Frequently Asked Questions

Jonas Helwig is a founder, CEO, entrepreneur and strategic advisor working across energy systems and artificial intelligence. He is co-founder of Kistora, an AI security company, and co-founder of Nuclearis Energy. His current work centres on energy storage, energy trading, grid balancing and AI systems built into the operations of energy businesses. He is also an author, currently writing a novel based on real events.

Four areas. Energy storage, covering both grid-scale batteries and sand batteries that store energy as heat. Energy trading across day-ahead, intraday and balancing markets. Grid balancing and the design of power systems that stay robust as generation becomes more weather-dependent. And AI systems for the operational side of energy businesses, CRM, telephony, dispatch and customer operations. Alongside that runs AI security work at Kistora and nuclear work at Nuclearis Energy.

He is co-founder of Kistora, which builds access control and governance for AI agents, and co-founder of Nuclearis Energy, which works in advanced nuclear.

Founder-led work and public commentary on energy systems, storage, firm power, grid reliability and nuclear, combined with hands-on software engineering. He writes about the point where computation and electricity meet, and he ships code in both domains rather than only commenting on them.

Kistora checks what an AI agent is about to do against policy before it acts, then allows, blocks or escalates it, and keeps a signed record of the decision. There is a fuller description on the about page and at kistora.ai.

Because agents have moved from producing answers to taking actions. Once an agent can run a deploy, move money or read a production database, output quality stops being the main risk and permission becomes the question. Controlling what an agent may do before execution, and keeping cryptographic proof of each decision afterwards, is a different problem from controlling what it says.

Two connections. He builds AI into working operational systems, CRM, customer operations, telephony and dispatch for energy businesses, and he works on AI security at Kistora. He also writes about AI as an infrastructure story, since large-scale compute is ultimately a question about electricity, grid capacity and physical build-out.

Yes. In July 2026 he published a blind, peer-judged benchmark in which six frontier models each solved the same six coding problems and then graded each other anonymously, with no model permitted to score itself. Opus 5 won all six domains with a mean peer score of 8.90 out of 10. The full method, heatmap and caveats are published on this site.

A thermal storage system that heats sand, or a similar cheap, abundant medium, with surplus electricity and holds that heat until it is needed. It is well suited to long-duration heat storage in industry and district heating, where the requirement is many hours of thermal output rather than the fast, short bursts of electrical power that lithium batteries provide. The two technologies solve different problems and are complementary rather than competing.

Because supply and demand must match second by second, and an increasing share of generation depends on weather. Storage moves energy across time, absorbing surplus when it is abundant and cheap, releasing it when it is scarce and expensive. That is what makes a volatile market manageable and a weather-dependent grid stable.

Buying and selling electricity across day-ahead, intraday and balancing markets. Day-ahead sets positions for the following day, intraday corrects them as forecasts change, and the balancing market handles the final seconds when actual demand differs from plan. Small forecast errors turn into real money quickly, which is why the software around forecasting, hedging and settlement matters as much as the physical assets.

Keeping generation and consumption matched continuously so that grid frequency stays within its tolerance. As weather-dependent generation grows, balancing increasingly depends on flexibility: storage that can charge or discharge on command, demand that can shift, and firm capacity that is available regardless of conditions.

Because nuclear supplies firm, high-density electricity that is available regardless of weather. Storage and flexibility solve part of the reliability problem, but a system also needs generation that is simply there. That combination, firm baseload plus storage plus flexibility, is what supports industrial capacity, energy security and long-term economic resilience.

Yes. He is co-founder of Nuclearis Energy, which works in advanced nuclear.

GOPBX is an open-source telephony platform he built: a FusionPBX-class PBX with a Go 1.25 control plane, a SvelteKit and shadcn-svelte admin frontend, and FreeSWITCH handling media and signalling. It provides REST CRUD across configuration objects, dynamic XML generation and ESL control, and is licensed MIT. The repository is at github.com/zendoj/GOPBX.

No. It is a functional clone rather than a fork and shares no code with FusionPBX. It targets the same problem space and a comparable feature surface, but the control plane is written from scratch in Go and the interface is a SvelteKit application rather than PHP. FreeSWITCH is deliberately kept as the media and signalling engine.

Svelte compiles the framework away, so there is no virtual DOM shipped to the browser and less work happens at runtime. Combined with Svelte 5's runes reactivity, that makes fine-grained live updates and aggressive client-side caching straightforward to build. He has published a measured case study of taking a data-heavy admin application from multi-second loads to sub-100 ms navigation.

DevLoop is an internal, AI-assisted development loop connecting in-app bug reporting, AI-proposed fixes against the real codebase, review in context and shipping. It removes the handoffs between issue tracker, editor, review tool and deploy pipeline rather than removing human judgement. It is not a public repository.

On GitHub at github.com/zendoj. The projects page on this site describes each one in more detail.

Yes. He is writing a novel based on real events surrounding the Aum Shinrikyo cult and the 1995 Tokyo subway attack. It is in research and draft. No publication date has been announced.

Energy storage, energy trading, grid balancing, firm power, energy density and energy security; AI infrastructure and AI-driven electricity demand; and software engineering, including frontend performance and telephony systems. Plus the novel, which is a separate thread.

Yes, one. He co-authored "Analysis of LSTM and GMDH network models for cryptocurrency forecasting", DOI 10.20998/2411-0558.2020.01.10. It is a single paper from 2020, not a research career.

A nuclear energy podcast he hosts with Adnan Belushi.

Yes. Public references include Swedish news coverage, podcast listings, international press, humanitarian fundraising visibility and public commentary on energy and infrastructure.

Swedish media reported on Jonas Helwig and his family in Odessa during the early phase of the full-scale war. A fundraiser was later organised to supply generators and connectivity equipment to people in need.

Yes. His background includes competitive hockey across several clubs, alongside a period in music production.

To be the canonical home for his biography, writing, projects, media references, publications and podcast activity, one authoritative source rather than a scatter of profiles.

Through the contact page, or via LinkedIn.