This weekend, Andrej Karpathy, the previous director of AI at Tesla and a founding member of OpenAI, determined he wished to learn a ebook. However he didn’t need to learn it alone. He wished to learn it accompanied by a committee of synthetic intelligences, every providing its personal perspective, critiquing the others, and finally synthesizing a closing reply below the steering of a "Chairman."
To make this occur, Karpathy wrote what he known as a "vibe code mission" — a bit of software program written rapidly, largely by AI assistants, meant for enjoyable quite than operate. He posted the consequence, a repository known as "LLM Council," to GitHub with a stark disclaimer: "I’m not going to help it in any approach… Code is ephemeral now and libraries are over."
But, for technical decision-makers throughout the enterprise panorama, wanting previous the informal disclaimer reveals one thing much more important than a weekend toy. In just a few hundred traces of Python and JavaScript, Karpathy has sketched a reference structure for probably the most vital, undefined layer of the trendy software program stack: the orchestration middleware sitting between company functions and the risky market of AI fashions.
As corporations finalize their platform investments for 2026, LLM Council gives a stripped-down have a look at the "construct vs. purchase" actuality of AI infrastructure. It demonstrates that whereas the logic of routing and aggregating AI fashions is surprisingly easy, the operational wrapper required to make it enterprise-ready is the place the true complexity lies.
How the LLM Council works: 4 AI fashions debate, critique, and synthesize solutions
To the informal observer, the LLM Council internet utility seems virtually an identical to ChatGPT. A person sorts a question right into a chat field. However behind the scenes, the applying triggers a classy, three-stage workflow that mirrors how human decision-making our bodies function.
First, the system dispatches the person’s question to a panel of frontier fashions. In Karpathy’s default configuration, this consists of OpenAI’s GPT-5.1, Google’s Gemini 3.0 Professional, Anthropic’s Claude Sonnet 4.5, and xAI’s Grok 4. These fashions generate their preliminary responses in parallel.
Within the second stage, the software program performs a peer assessment. Every mannequin is fed the anonymized responses of its counterparts and requested to judge them based mostly on accuracy and perception. This step transforms the AI from a generator right into a critic, forcing a layer of high quality management that’s uncommon in commonplace chatbot interactions.
Lastly, a chosen "Chairman LLM" — at present configured as Google’s Gemini 3 — receives the unique question, the person responses, and the peer rankings. It synthesizes this mass of context right into a single, authoritative reply for the person.
Karpathy famous that the outcomes have been usually shocking. "Very often, the fashions are surprisingly prepared to pick one other LLM's response as superior to their very own," he wrote on X (previously Twitter). He described utilizing the software to learn ebook chapters, observing that the fashions persistently praised GPT-5.1 as probably the most insightful whereas score Claude the bottom. Nevertheless, Karpathy’s personal qualitative evaluation diverged from his digital council; he discovered GPT-5.1 "too wordy" and most well-liked the "condensed and processed" output of Gemini.
FastAPI, OpenRouter, and the case for treating frontier fashions as swappable parts
For CTOs and platform architects, the worth of LLM Council lies not in its literary criticism, however in its development. The repository serves as a major doc exhibiting precisely what a contemporary, minimal AI stack seems like in late 2025.
The applying is constructed on a "skinny" structure. The backend makes use of FastAPI, a contemporary Python framework, whereas the frontend is a regular React utility constructed with Vite. Knowledge storage is dealt with not by a fancy database, however by easy JSON recordsdata written to the native disk.
The linchpin of the complete operation is OpenRouter, an API aggregator that normalizes the variations between numerous mannequin suppliers. By routing requests via this single dealer, Karpathy prevented writing separate integration code for OpenAI, Google, and Anthropic. The applying doesn’t know or care which firm offers the intelligence; it merely sends a immediate and awaits a response.
This design alternative highlights a rising pattern in enterprise structure: the commoditization of the mannequin layer. By treating frontier fashions as interchangeable parts that may be swapped by enhancing a single line in a configuration file — particularly the COUNCIL_MODELS listing within the backend code — the structure protects the applying from vendor lock-in. If a brand new mannequin from Meta or Mistral tops the leaderboards subsequent week, it may be added to the council in seconds.
What's lacking from prototype to manufacturing: Authentication, PII redaction, and compliance
Whereas the core logic of LLM Council is elegant, it additionally serves as a stark illustration of the hole between a "weekend hack" and a manufacturing system. For an enterprise platform group, cloning Karpathy’s repository is merely step considered one of a marathon.
A technical audit of the code reveals the lacking "boring" infrastructure that industrial distributors promote for premium costs. The system lacks authentication; anybody with entry to the online interface can question the fashions. There isn’t any idea of person roles, which means a junior developer has the identical entry rights because the CIO.
Moreover, the governance layer is nonexistent. In a company atmosphere, sending information to 4 totally different exterior AI suppliers concurrently triggers quick compliance issues. There isn’t any mechanism right here to redact Personally Identifiable Info (PII) earlier than it leaves the native community, neither is there an audit log to trace who requested what.
Reliability is one other open query. The system assumes the OpenRouter API is all the time up and that the fashions will reply in a well timed style. It lacks the circuit breakers, fallback methods, and retry logic that preserve business-critical functions operating when a supplier suffers an outage.
These absences should not flaws in Karpathy’s code — he explicitly said he doesn’t intend to help or enhance the mission — however they outline the worth proposition for the industrial AI infrastructure market.
Corporations like LangChain, AWS Bedrock, and numerous AI gateway startups are primarily promoting the "hardening" across the core logic that Karpathy demonstrated. They supply the safety, observability, and compliance wrappers that flip a uncooked orchestration script right into a viable enterprise platform.
Why Karpathy believes code is now "ephemeral" and conventional software program libraries are out of date
Maybe probably the most provocative facet of the mission is the philosophy below which it was constructed. Karpathy described the event course of as "99% vibe-coded," implying he relied closely on AI assistants to generate the code quite than writing it line-by-line himself.
"Code is ephemeral now and libraries are over, ask your LLM to vary it in no matter approach you want," he wrote within the repository’s documentation.
This assertion marks a radical shift in software program engineering functionality. Historically, corporations construct inner libraries and abstractions to handle complexity, sustaining them for years. Karpathy is suggesting a future the place code is handled as "promptable scaffolding" — disposable, simply rewritten by AI, and never meant to final.
For enterprise decision-makers, this poses a troublesome strategic query. If inner instruments might be "vibe coded" in a weekend, does it make sense to purchase costly, inflexible software program suites for inner workflows? Or ought to platform groups empower their engineers to generate customized, disposable instruments that match their actual wants for a fraction of the price?
When AI fashions choose AI: The harmful hole between machine preferences and human wants
Past the structure, the LLM Council mission inadvertently shines a light-weight on a particular threat in automated AI deployment: the divergence between human and machine judgment.
Karpathy’s commentary that his fashions most well-liked GPT-5.1, whereas he most well-liked Gemini, means that AI fashions could have shared biases. They may favor verbosity, particular formatting, or rhetorical confidence that doesn’t essentially align with human enterprise wants for brevity and accuracy.
As enterprises more and more depend on "LLM-as-a-Decide" programs to judge the standard of their customer-facing bots, this discrepancy issues. If the automated evaluator persistently rewards "wordy and sprawled" solutions whereas human clients need concise options, the metrics will present success whereas buyer satisfaction plummets. Karpathy’s experiment means that relying solely on AI to grade AI is a method fraught with hidden alignment points.
What enterprise platform groups can study from a weekend hack earlier than constructing their 2026 stack
Finally, LLM Council acts as a Rorschach take a look at for the AI trade. For the hobbyist, it’s a enjoyable method to learn books. For the seller, it’s a risk, proving that the core performance of their merchandise might be replicated in just a few hundred traces of code.
However for the enterprise expertise chief, it’s a reference structure. It demystifies the orchestration layer, exhibiting that the technical problem is just not in routing the prompts, however in governing the information.
As platform groups head into 2026, many will probably discover themselves watching Karpathy’s code, to not deploy it, however to grasp it. It proves {that a} multi-model technique is just not technically out of attain. The query stays whether or not corporations will construct the governance layer themselves or pay another person to wrap the "vibe code" in enterprise-grade armor.
