Aiolum Browse stories
All stories
AI coding assistant for large production codebases · Not stated in the sources

Jolt AI: Revenue, sale and how it was built

Small teamFundedExitB2BDeveloper tools

Jolt AI (AI coding assistant for large production codebases, Not stated in the sources): Team acquired by Ramp, October 2025. Evidence grade E: no revenue figure was checked. Team: Founder and CEO Yev Spektor, a CTO and a principal engineer: 3 people when Ramp hired the whole team in October 2025. Launched 23 January 2025; sold 9 months after launch.Added 8 November 2026

Outcome
Team acquired by Ramp, October 2025
Evidence
E no revenue figure was checked
Launch
23 January 2025
Status
Sold, 9 months after launch
Team
Founder and CEO Yev Spektor, a CTO and a principal engineer: 3 people when Ramp hired the whole team in October 2025
Outside funding
Just under US$2.5M since 2022 from investors including Alumni Ventures, 8-Bit Capital and Engineering Capital, as Crunchbase News reported
Main channel
Not stated
No revenue figure

What is not verified

No revenue figure, number of paying teams or sale price was found. The sale and the team size come from Ramp: its co-founder's statements to Crunchbase News and its own blog post of 6 October 2025. The just-under US$2.5 million in funding is reported by Crunchbase News only. The customer quotes on the 2024 site, the SOC 2 Type II badge (no report was linked) and the codebase sizes are the company's own statements. Looked for: a payment-processor profile, a founder interview with figures, an earlier copy of the pricing page (the Internet Archive answered 429 and was not asked again). A reader could look for an interview with Yev Spektor about the business before the sale. This story is grade E: the business is real and its product, price and customers are described from public pages, but no revenue figure was checked. Aiolum shows it as an idea worth studying, not as proof that it makes money.

Jolt AI was an AI coding assistant, a program that writes and explains code from a request in plain language, built for very large codebases. Its key part was context retrieval: finding the few files out of thousands that a language model must read to answer a request. Founder and CEO Yev Spektor wrote that most tools use vector-embedding search for this, which works badly on code, and that Jolt used its own search instead.

The company started in 2022 as Multiple, a load-testing platform (software that sends many simulated users to a service to see when it breaks), and raised just under US$2.5 million, Crunchbase News reported. In 2024 it pivoted to Jolt; the public beta opened on 23 January 2025, and by July 2025 the pricing page listed a Team plan at US$200 per seat a month. On 6 October 2025 Ramp hired all three people in an acqui-hire, a purchase made for the team rather than the product; Crunchbase News reported that the product was not part of the deal and that the terms were not disclosed.

No revenue figure, customer count or sale price was found, so this story carries grade E: the headline is the sale of the team itself, confirmed by Ramp's own blog. The usejolt.ai domain no longer resolves.

Jolt AI: Team acquired by Ramp, October 2025. AI coding assistant for large production codebases, Not stated in the sources, team of 2 to 10, evidence grade E.
The story in one card. Share it with the address of this page.
01

How much did Jolt AI make, and what was it sold for?

Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.

A three-person company that pivoted from load testing to an AI coding assistant and was hired as a team by Ramp nine months after its public beta; no revenue or price for the deal was published
FactValueGradeHow we knowSources
SaleRamp hired the whole three-person team on 6 October 2025BCrunchbase News, 6 October 2025, from Ramp's co-founder and CTO Karim Atiyeh: Ramp "has acquired the team of a three-person startup called Jolt AI". Ramp's own blog of the same day says it is "excited to welcome the Jolt team to Ramp".[1, 2]
What was boughtthe team only, not the productBCrunchbase News: "Ramp's purchase, notably, involved only the company's three-person team, and not its product." The team joined Ramp's engineering platform team.[1]
Sale pricenot disclosedBNeither Crunchbase News nor Ramp's blog gives an amount.[1, 2]
Company start2022, as Multiple, a load-testing platformCCrunchbase News gives a "2022 inception" and the load-testing start. The September 2023 copy of multiple.dev calls it "a developer-centric load test platform" under the same company name, PRL Inc., that is printed on usejolt.ai. The founder's dev.to profile, opened in December 2022, says "Founder @ multiple.dev".[1, 7, 6]
Pivot to AI2024CCrunchbase News, from the founder. The oldest Internet Archive copy of usejolt.ai, 19 September 2024, already shows the coding assistant with quotes from named customers.[1, 3]
Public beta23 January 2025CFounder's post on dev.to: "The Jolt public beta is live today, and sign-ups are now open for the first time." Used as the launch date on this page.[6]
Team at the sale3 people: the founder and CEO, a CTO and a principal engineerBCrunchbase News names all three. Only Yev Spektor is called founder in the sources read.[1]
Outside fundingjust under US$2.5 million since 2022CReported by Crunchbase News, with Alumni Ventures, 8-Bit Capital and Engineering Capital named as investors. No round dates were found.[1]
Paid plan, July 2025Team at US$200 per seat a month, minimum 5 seatsCPricing page, Internet Archive copy of 8 July 2025. A free trial and a custom Enterprise plan were also listed. No count of paying teams was found.[4]
Largest codebase in useover 8 million lines in January 2025; over 10 million lines by July 2025CThe founder's launch post and the pricing page. Not checked by anyone else.[6, 4]
Product statusdomain usejolt.ai does not resolve, 8 November 2026CChecked on this page with a DNS lookup. The last Internet Archive copy of the home page is from 15 September 2025, three weeks before the sale. No shutdown notice was found.[5]
Time from public beta to the sale9 monthscalcCalculated: January 2025 to October 2025.[6, 1]

How to read this table: a grade A or B value comes from the buyer, in its own blog or in its statements to Crunchbase News. A grade C value is the company's own page, read in an Internet Archive copy, or the founder's own post. A "calc" value was computed on this page. No revenue figure or sale price was found, so the headline is the sale itself.

Jolt AI plans in July 2025, from an Internet Archive copy of the company's pricing page of 8 July 2025 (grade C)
PlanPer month, billed yearly (US$)Per month, billed monthly (US$)Limit
Trial00500 credits; repositories up to about 120,000 lines
Team200 per seatminimum 5 seats; unlimited usage under a fair-use ceiling; any repository size; context from several repositories
Enterprisecustomusage reports, training workshops, single sign-on, role-based access, on-premise option

The page said the price was all-inclusive: the customer did not pay the model providers separately, and the paid plans had a credit ceiling set as a fair-use policy. Fewer than 5 Team seats were possible on request by email. The page did not show a yearly price.

02

When did Jolt AI launch, and how did it grow?

From a load-testing service to an AI coding assistant to a team inside Ramp, in about three years. Each entry carries its grade.

  1. 2022CThe company starts as Multiple, a load-testing service for developers, and later raises just under US$2.5 million in total, as Crunchbase News reported. [1, 7]
  2. 2024CIt pivots to Jolt, an AI coding assistant for large codebases. By 19 September 2024 the site shows quotes from engineering leaders at named companies. [1, 3]
  3. 23 January 2025CPublic beta: sign-ups open for the first time. The founder writes that the largest codebase using it is over 8 million lines. [6]
  4. 8 July 2025CThe pricing page lists a Team plan at US$200 per seat a month with a minimum of 5 seats, and states that it uses Google, Anthropic and OpenAI models. [4]
  5. 6 October 2025BRamp hires the three-person team. The CEO of an early Jolt customer had introduced them, Crunchbase News reported. The product is not part of the deal and the terms are not disclosed. [1, 2]
03

Why did Jolt AI work?

Four causes, in the order they mattered. Each one is a reusable lesson.

01

Own search instead of embedding search

The founder wrote that vector-embedding search, the usual way AI coding tools pick files, loses accuracy as a codebase grows. Jolt built its own search, which it called HyperContext, over a connection to the customer's git provider. The pricing page said this search is compute-intensive, which is why Jolt cost more than other tools.

02

A plan before the code

In 2024 Jolt first wrote an editable file-by-file plan for a task, and only then the code. A developer could change the plan, and could ask for revisions by leaving a review comment, as on a pull request.

03

Every editor and one price

It ran as a VS Code extension (which also covers Cursor and Windsurf), a JetBrains plugin, a desktop app for other editors and a web app. The pricing page stated that model costs were included in the seat price, so a team had one bill.

04

Engineers as the asset

Ramp's co-founder told Crunchbase News that the deal was about the team, and the product did not change hands. What the team had learned about search in large codebases was what Ramp wanted for its own engineers.

How it was built

These points are the company's own statements on its pages, read in Internet Archive copies, and in the founder's launch post. They are grade C. [3, 4, 5, 6]

  • Context search: a proprietary engine, HyperContext, that searched, navigated and read the codebase through a connection to GitHub or another git provider, and picked the files for each request.
  • Models: a combination of Google, Anthropic and OpenAI models, with Jolt choosing the model for each step; customers could also bring their own model API keys.
  • Clients: a VS Code extension, a JetBrains plugin, a desktop app and a web app, all on one account.
  • An API that turned a text or image prompt into a pull request; the pricing page said teams connected it to Linear, Jira, Sentry and GitHub Actions.
  • Security: the site claimed a SOC 2 Type II audit and said that neither Jolt nor its model providers trained on customer code.

Tools named in the sources

Each one opens its page in the Tools tab, with every story that uses it.

04

Who tried this before Jolt AI, and why did they fail?

Three developer-tool companies in our database of failed startups promised to make developers faster or help them understand code, and failed against free tools or tools bundled by larger companies. The values below are from that database and are not independently verified.

Died 2014 to 2024Burned US$21M

Kite DevTool

The promise

AI code completion inside the editor, years before GitHub Copilot.

Why it died

It could not make its free product pay, and GitHub Copilot, built on OpenAI models and given GitHub's reach, took the market.

Competition
Died 2020 to 2024Burned US$10M

CodeSee

The promise

Interactive maps that help developers understand a large codebase.

Why it died

Engineering leaders could not see a clear return, and teams used free alternatives for a problem that came up only now and then.

No Market Need
Died 2015 to 2018Burned US$120K

Omniref

The promise

A Wikipedia for code, where developers explain and annotate source code.

Why it died

Without a community or money it lost to GitHub and Stack Overflow, and AI tools later took over the task.

Competition

The failed startups come from our database of 1,749 dead startups, retold here in our own words. Its entries are AI-assisted summaries of public sources, so these cards carry no evidence grade.

What changed

Kite, CodeSee and Omniref each tried to help developers write or understand code before large language models could read code well. Kite lost to a tool bundled by GitHub, and CodeSee could not show a return large enough for a budget. Jolt came after the models, so its work was the search around the model, not the model.

The end looks similar in one way: a small team with a specialised product faced Cursor, GitHub Copilot and the model makers' own coding agents. Jolt did not fail in the database's sense. A large company that wanted to make its own engineers faster bought the team, and the product stopped.

05

Can one person copy Jolt AI now?

The product, hard. The route to a buyer, yes.

A search engine for multi-million-line codebases needed experienced engineers, outside money and enterprise security work, and it competed with tools from very large companies. The route is open to a small team: solve one hard problem in how engineers work, sell it to engineering teams, and become known to companies that want that skill for their own staff.

Where it is still open: Search, review and testing tools for very large codebases, legacy languages and code spread over many repositories, where general coding tools still need the developer to supply the context.

  1. 1Pick a problem that the large AI coding tools handle badly, such as very large or old codebases.
  2. 2Show it on real customer code, with named engineering leaders who will say what changed.
  3. 3Price per seat with the model cost included, so a team gets one predictable bill.
  4. 4Treat a buyer who wants the team as a possible outcome from the start.

What could go wrong for Jolt AI?

The risks are the mirror image of the causes above.

What is not verified

No revenue figure, number of paying teams or sale price was found. The sale and the team size come from Ramp: its co-founder's statements to Crunchbase News and its own blog post of 6 October 2025. The just-under US$2.5 million in funding is reported by Crunchbase News only. The customer quotes on the 2024 site, the SOC 2 Type II badge (no report was linked) and the codebase sizes are the company's own statements. Looked for: a payment-processor profile, a founder interview with figures, an earlier copy of the pricing page (the Internet Archive answered 429 and was not asked again). A reader could look for an interview with Yev Spektor about the business before the sale.

A team sale, and the product stopped

Ramp bought the team, not the product, as Crunchbase News reported. The last archived copy of usejolt.ai is from 15 September 2025, and on 8 November 2026 the domain did not resolve. No notice to customers was found. This page counts the deal as an exit because Ramp told Crunchbase News that it had acquired the team, and the article calls the deal a purchase.

Competition

The pricing page compared Jolt with Cursor and GitHub Copilot and said some teams used them side by side; on this site, Conductor (bought by Meta) and ZeroEntropy (bought by Notion) are other small AI developer-tool teams that ended in a sale to a larger company.

Dependence on model providers

The pricing page of July 2025 said Jolt used Google, Anthropic and OpenAI models and that its seat price included model costs. So a change in those providers' prices or terms fell on Jolt's margin, not on the customer's bill.

Sources

The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.

  1. B Crunchbase News, 6 October 2025: Exclusive: Fintech Decacorn Ramp Acquires Jolt AI to Help Its Engineers 'Build Faster', by Marlize van Romburgh
    Read in full (downloaded 7 November 2026 by the pipeline). Ramp's co-founder and CTO is quoted. Source for the team sale, the team of three, the product not being part of the deal, the 2022 start, the pivot and the funding.
  2. A Ramp blog, 6 October 2025: Scaling Ramp's AI platform with the Jolt team, by Karim Atiyeh
    Read in full. The buyer's own statement that the Jolt team joined Ramp, with a quote from Yev Spektor as Jolt's founder and CEO.
  3. C Jolt AI home page, Internet Archive copy of 19 September 2024
    Read in full. Source for the product in 2024, the plan-then-code workflow and the named customer quotes.
  4. C Jolt AI pricing page, Internet Archive copy of 8 July 2025
    Read in full. Source for the plans and prices, the model providers, the HyperContext search, the API and the comparison with Cursor and Copilot.
  5. C Jolt AI home page, Internet Archive copy of 15 September 2025
    Read in full. The last archived home page before the sale: IDE support, multi-repository context, the SOC 2 claim. The live domain did not resolve on 8 November 2026.
  6. C DEV Community, 23 January 2025: Introducing Jolt: AI Codegen and Chat for 100K to Multi-Million Line Codebases, by Yev
    Read in full. The founder's launch post for the public beta, also published on usejolt.ai. The author profile on the page says "Founder @ multiple.dev".
  7. C Multiple home page, Internet Archive copy of 19 September 2023
    Read in full. The load-testing product before the pivot, under the company name PRL Inc.
  8. C Database of failed startups, local export
    Local export. AI-written summaries, not independently verified. Source for the failed predecessors.

Run this company? Send corrections or a verified figure to aiolum@livyden.com

Back to all stories