Want a smart insight into your inbox? Sign up for our weekly newsletters to get the only thing that is important to enterprise AI, data, and security leaders. Subscribe now
Traditionally, the product’s release can be burdened, which requires several sign office, endless tinkering, bureaucracy and friction points.
Jennpark Have developed a very different approach.
The AIwork Space Company’s lean team practices AI-Local work-or ‘to work’, if you do-so that they can move forward with what they call the “gene speed”. With this, they can rapidly release new products and features in firefighters (almost every week or more), which lead to permanently repeated revenue (ARR). As if The company makes pride“This can be the fastest growing starting so far in terms of Arr.”
“When people are operating in a local way, everyone is basically a manager,” Kahwa (K) Zhu, co-founder and CTO told Venture Bat. “They are equipped with a team of AI agents, who are like their reports, and they are in accordance with their ability, and bring this feature from the end to the end.”
AI scaling is hit by its limits
Power caps, increasing token costs, and delays in reduction are changing enterprise AI. Join our special salon to find out how high teams are:
- Changing energy into a strategic advantage
- Effficient carcincing effectively of real throptic benefits
- Unlocking competitive ROI with a sustainable AI system
Save your space to stay ahead:
Aggressive rollout, stoking competition
JennparkLaunched in June 2024 ManphonicWas initially focused on the search for AI. But despite reaching an impressive 5 million users, the company turned away from this initial product Super AgentInstead of following the static continuity of the traditional search -like measures, the job lies the best tools or all agents, adjusting the results of the gauge and in real time.
Launching on April 2, the Super Agent has been strengthened through an anthropic cloud and can reduce the White Caller Office work by 5 minutes a afternoon. For example, it can make calls, downloads, facts testing, producing podcasts, producing draft documents, performing deep research, and can draw spreadsheets and slides together.
“We still see it as a kind of search, but it’s technically more developed,” said Zhu, who has more than 20 years of experience in search of Google and Bedo.
The company has aggressively added more and more features in the last four months. Her rollout and milestone a way is:
- April 11: Million reached 10 million ARR just after the Super Agent launch
- 22 April: Introduced AI Slides (Full of Hundreds of Templates)
- April 28: Developed a Personal Super Agent with Inclusive Personas
- May 2: Million Kill 22 million RARs, Launch Acquisition After a month
- May 8: Rolled AI Sheets that prepare a full spreadsheet in one click
- May 15: A full agent download agent and AI drive that manages and stores files
- May 19: Million kill 36 million ARR
- May 22: Phone Calling AI rolled out
- June 4: Introducing an AI Secretary that manages Gmail, Calenders and Google Drive
- June 10: An AI browser and MCP store developed that features expansion browsing capabilities and a tool market.
- June 18: Introducing AI documents for the creation and management of the document
- June 25: Introduced Studio Design with “Kinawa” capabilities to create visual content
- July 10: Eliminated AI Pls to make podcasts with simple indicators
- July 17: Advanced Editing features introduced for AI slides
- July 31: Rolled AI Slides 2.0
- August 1: Introducing Multi Agent Orchestation that can manufacture 10 agents simultaneously
Genspark is also heating the AI agent’s space with a friendly competition. Openi after his announcement Chat GPT Agent In mid -July, Guinness Spark compared comparatively and is “much confident” in the ability to maximize the competitors. To drive this point home, the company “launched”$ 1 million by -back AI Shurdown“Challenging consumers is to prevail where other platforms improve Ginus Spark Super Agent.
In the first phase, consumers were assigned a 12 -page financial slide to make a 12 -page financial slide using Genpack and Chat GPT agents. Consumers identified 429 cases, where the latter performed well to the former, each earned $ 100 for his efforts.
In Round 2 (which ended on Monday, August 4), Guinness Spark had already won and opened the competition from any AI tool as an opponent. Consumers were challenged to use exactly the same gesture to make the slide on the Gins Spark and their selected AI tools, then upload them to Gemini for diagnosis.
The company “tried to start any drama here – just really passionate posted on x. “This shows that we are all moving the boundaries in the right direction.”
Some user’s reaction:
How GensPark’s AI ancestral team vibs
The secret of Guinness Spark is its lean, AI-local team and engineering philosophy that is “low control, more tools”. Zhu explained that more than 80 Than AI of its code wrote by AI, which does not codify per second, “because web coding type indicates that you never see the code.” Rather, Jeans Spark has a “very strict” code review process to help guarantee the quality of the code base.
“We just need a very small AI-local team to work in a kind of superhero format, such as Revenge“Zhu said, who said he would slowly include team members as needed.” AI coding and AI workflow is so powerful, it is important. “
He said that today’s enterprise teams should be reorganized in a completely different way. They have arranged a thousand teams with different levels of management and have seen how office politics can introduce friction.
On the contrary, Gunis Spark’s team communicates in a “very transparent manner”, and the production capacity is “extremely high”. “Everyone is working on a product that can ship,” said Jho. “I am sure it will be normal, because AI is actually helping more people do their job better.”
He also emphasized the importance of drowning himself into his product. From self -designers to marketing team, “We really eat our own dog food. We are our own product user. Likewise we will continue to improve experience.”
Inside the flagship super agent of the Guins Spark
Zhu noted that, until the launch of the Palaxy in December 2022, he gave rise to excitement about the AI’s ability to change the search. Still, he followed the hard work flu, with the platform:
- Analyze the questions and expand the keywords.
- Recover high web results.
- Re -work/summarize for the final response.
It was sufficient for the basic things Adequate, but “fell” in complex scenes such as technical comparisons, deep research and multi -phase and multi -factor purchase. Zhu said, “In essence, it was just like trying to visit the maze with a fixed turn.
Guinness Spark created its search engine at a similar foundation, which includes cross -checking of controversial agents to confirm special data sources, deepening of complex questions, and verify very complex statements for “sharp, flight”. According to Zhu, but they realized that they were still “fibers” through the default, default workflow.
Super Agent Nine Different Sizes, Different Uses Different Large models of tongue (LLMS) Compound off agents (MOE) system. Models work in stages, assign on the basis of feature and strength, then cross each other. Super Agent is more than 80 tools (equipped with sub -agents that can produce azar code that can make autonomous phone calls) and more than 10 datases made from web, partners and reserves.
Zhu explained, “Gunin Spark Claude, Openi, Google Gemini, Dipoic, AI’s Grick 4 and others,” then we allow each one to produce our production, and we have a collective model to see and analyze the results which process is extremely effective. “” In this way, we improve accuracy, reduce deception. ”
The company also fixes its frontier model. However, they are not excessive aggressive about creating the latest system like DPCC V3 or V4, Zhu emphasized. The purpose is to perform the low -level but heavy lifting work.
“We are not trying to advance the boundaries of the Frontier Model,” he said. “We are trying to reduce costs and delays, because many proprietary models are very big, very slow and very expensive that are very expensive for relatively simple easy tasks.”
As far as coding trends, Ganes Spark’s goal is to allow everyone to experience, even for non -programmers where this concept can be a “far away”.
Zhu said, “Many people think, ‘Web coding, I’ve heard of it, it sounds good, but I’m not familiar with the integrated developer environment, I am not familiar with the code.” “Using jeans spark, people can actually wear.”
