Arcfront Launches, Says 4 of 5 Frontier Labs Buy Its Data

Arcfront, a startup that turns records of real expert work into training data for AI models, came out of stealth on 20 November and said four of the five frontier AI labs already work with it, per its launch post on X.

The company also claims more than 300 enterprise partners and more than 1,000 RL environments, the practice tasks AI models are trained and tested on. It names none of the labs or partners.

Business Pill: The Cost of Checking

An evergreen Business Engineer explainer. It teaches the general idea only and does not discuss Arcfront.

The key insight: Arcfront came out of stealth claiming four of the five frontier AI labs as customers, more than 300 enterprise partners and more than 1,000 RL environments, none named; a relayed $1 billion revenue figure is not in its own materials (Arcfront; Wall St Engine).

What Arcfront Says It Does

On its website, Arcfront says it “studies how real work gets done and turns that record into training data for frontier models.”

It calls the product “Decision Arc Data: every turn on the path from situation to outcome,” and says it builds that into RL environments and evals for frontier labs.

The site lists two product lines: RL environments for software, site reliability, cyber and co-work tasks, and sector evals built from workflows at hospitals, banks, law firms and engineering teams.

In a reply on X, co-founder Anshuman Singh said the company uses “patterns of the work they’ve already done and documented in the institutions,” giving radiology reports and the radiologists’ assessments attached to them as an example.

Arcfront says four of the five frontier labs are customers but names none of them (Arcfront).
Arcfront says four of the five frontier labs are customers but names none of them (Arcfront).

Who Is Behind It

Singh announced the launch with co-founder Naman Bhalla and “deep support” from the account @asxna. He wrote that in under a year, “four of the five frontier labs became customers.”

Singh co-founded the Indian coding-education company InterviewBit, later Scaler, in 2015; InterviewBit raised $20 million in 2020, per TechCrunch. He wrote that he competed in two ICPC World Finals, the top university programming contest.

The website lists six open roles, four in San Francisco and two in Bengaluru, and its footer names the company as Evaratus Inc.

Bar chart: Arcfront lists four open roles in San Francisco and two in Bengaluru
Arcfront lists six open roles across two offices (arcfront.ai).

The Numbers

Four of five is 80% of the frontier labs, by Arcfront’s own count (our arithmetic). Arcfront does not say which five labs it counts as frontier, and no lab has confirmed it is a customer.

The Structural Read

As we read it, the pitch is that the scarce input is no longer text on the internet but records of how experts make decisions at work.

As we read it, the claims will matter only if a lab or partner confirms them; for now they are the company’s own.

Arcfront website

“Arcfront studies how real work gets done and turns that record into training data for frontier models.”

Three Implications

FOR AI LABS Arcfront says four of the five frontier labs already work with it, without naming them (Arcfront).

FOR ENTERPRISES the company says it has more than 300 enterprise partners whose work records feed its data (Arcfront).

FOR VERIFICATION a relayed $1 billion revenue figure is not in Arcfront’s own launch materials (Wall St Engine; Arcfront).

What Is Not Established

A widely shared post from Wall St Engine said Arcfront reached $1 billion of annual recurring revenue with under $2 million raised. Those figures do not appear in Arcfront’s launch post, Singh’s post or the company website as we read them, so we do not report them as company claims.

Arcfront has not disclosed funding, investors, a valuation or revenue in the materials we read, and the customer claims are its own.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The AI Supercycle

The Business Engineer map of who builds, finances and captures value in AI.

Read the Map of AI →

The Bottom Line

Arcfront launched claiming four of the five frontier labs, more than 300 enterprise partners and more than 1,000 RL environments, all unnamed and unconfirmed by the labs, on a pitch that real workplace records can teach AI models judgment.

97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Arcfront’s claims are from its own launch post on X, co-founder Anshuman Singh’s posts and the company website as read on 21 November. None of the labs or partners is named, and we found no lab confirmation. The $1 billion revenue figure is from a Wall St Engine post and is not in the company’s own materials. InterviewBit’s founding and 2020 raise are from public company records and TechCrunch. The 80% is our arithmetic. Arcfront was not contacted for this piece.

Sources: Arcfront on X, 20 Nov 2026 18:40 UTC: launch post — four of five frontier labs, 300+ partners, 1,000+ RL environments · Anshuman Singh on X, 20 Nov 2026 18:37 UTC: launch with Naman Bhalla; four of five frontier labs became customers in under a year · Arcfront website (read 21 Nov 2026): Decision Arc Data, RL environments, sector evals, six open roles, Evaratus Inc. · Anshuman Singh reply on X, 20 Nov 2026: using documented work patterns, e.g. radiology reports · Wall St Engine on X, 20 Nov 2026: relayed $1B ARR and under $2M raised (not in company materials) · TechCrunch, 27 Jan 2020: InterviewBit secures $20M

Scroll to Top
FourWeekMBA