AI has become the strategic core of enterprises. It is not only a topic tool, but also a productivity engine. As Fortune 500 companies try to integrate AI collaboration into their businesses, enterprise AI is ushering in a new era of competition. However, unlike the past SaaS model, to win in this change, AI startups must not only develop quickly, but also be accurate and have a long tail.
Well-known venture capital firm a16z (Andreessen Horowitz) published a long article titled " From Demo to Transaction: Insights into Building Enterprise-Grade AI " to analyze how AI startups can stand out from the competition. The following is a compilation by ABMedia.
Table of Contents
ToggleIt’s not hard to show off your skills, but it’s hard to create AI products that can last for a long time
The wave of generative AI has made many products like " GPT wrappers " , which are often easy to copy and will eventually be swallowed by large models. In reality, it is easy to create a stunning demo that can trigger topics, but making AI work stably in real enterprise scenarios is the most difficult key difference.
From model selection, evaluation criteria, task and workflow design, to deeply embedding products into customer data and business logic, AI startups face extremely tedious and variable engineering challenges.
a16z emphasized that enterprises not only require accuracy, but also care about the reliability and liability risks of models in key areas such as law or accounting. Therefore, if AI wants to enter the core system of an enterprise, it cannot rely solely on surface effects, but must also be long-tail and fault-tolerant.
AI startup competition heats up: 10x annual revenue growth becomes the new threshold
In the past, reaching $1 million in annual recurring revenue (ARR) was considered an early success indicator, but today this is lower than the average for AI startups. According to Stripe and market data, many AI companies have achieved more than 10 times the annual revenue growth, and some have even exceeded millions of revenue in a few months.
a16z believes there are two reasons behind this:
First, many companies have set up budgets dedicated to AI and are proactively looking for solutions, so the purchasing process is quick and the budget is sufficient. Second, AI software is usually sold as the work results themselves, rather than selling auxiliary tools to help people get their work done.
In other words, AI software will directly replace human labor to complete work, and will shift to eating up the " human labor budget " rather than the "traditional IT budget", and the value of such contracts will therefore be improved overall.
The popularization of technology has led to an explosion of AI applications: edge scenarios have also been seen
At the same time, AI model pricing is falling rapidly, with the price of OpenAI's o3 model falling by as much as 80% in one year. Coupled with the popularity of development tools such as Replit and Cursor, non-technical users can also build AI applications, greatly accelerating the explosion of AI products.
This means that there will be a surge in personalized travel planning assistants, health monitoring dashboards, and internal tools to solve corporate business processes:
AI is giving rise to many small niche markets and edge application scenarios that had little economic benefits and could not be processed on a large scale in the past, allowing more innovative products to be implemented.
The fastest winner: first mover advantage is the key
AI buyers are not only overwhelmed by information, but also have long shown brand preferences. a16z emphasized that "Are you the first AI solution to enter this field?" has become the core of whether many companies will buy it.
New startups such as Cursor , an IDE agent, and Harvey , a favorite among legal professionals, have become representative products of specific categories by relying on rapid iteration and user penetration in the early stages. This speed not only outperforms traditional giants, but also marginalizes competitors before they have time to react. In the AI battlefield, speed is monopoly power.
To go far, you still need to build a moat. AI itself is not a panacea.
AI capabilities do not equal business barriers. In order to avoid being replaced by faster runners, a16z reminds successful startups that they need to build a real "moat":
Become the data system hub: Like Eve or Salient , they start with voice or unstructured data and gradually become the key data system of the industry.
Establish usage habits and bind them to workflows: For example, Decagon strengthens the human-machine collaboration process, making it difficult for users to disconnect.
Deeply integrate the enterprise ecosystem: Companies such as Tennr and HappyRobot connect old internal medical and logistics systems to provide irreplaceable value.
Establish trust and cooperative roles: Excellent AI vendors are no longer just tool suppliers, but also partners who help companies formulate AI strategies.
From fake popularity to real strength, AI startups seek more accuracy and depth
The AI boom in recent years has created golden opportunities for global companies, but it has also made the competition extremely cruel. AI startups that survive must not only iterate quickly, but also be able to accurately target demand and create a solid moat.
From demos to deals, from ideas to products, from simple tools to partnerships, the road to AI startups is much longer than imagined, and it adds more value to the teams that devote themselves to it.
Risk Warning
Cryptocurrency investment carries a high degree of risk. Its price may fluctuate drastically and you may lose all your capital. Please assess the risk carefully.
After the explosion of LLM AI such as ChatGPT, the development of voice interaction products has remained stagnant. ElevenLabs' latest voice AI assistant 11.ai attempts to break this deadlock. It can not only communicate with people in natural voice, but also integrate work tools and perform practical tasks, from scheduling to writing customer summaries, enhancing work efficiency and life experience.
Table of Contents
ToggleMore than just Siri: 11.ai helps you complete daily tasks automatically
For many years, voice assistants from Siri to Alexa have been focusing on "controlling technology with words", but most of them can only answer questions, turn lights on and off, or play music. ElevenLabs' 11.ai emphasizes that AI assistants are no longer just chat partners, but real assistants that can penetrate people's workflows and complete multi-step tasks.
The official announcement pointed out that users only need to speak out their needs, such as " Help me organize today's schedule and add it to Linear ", " Help me use Perplexity to check the background information of the client I have a meeting with today ", and 11.ai can connect multiple tools to automatically complete a series of actions such as task creation, data search and information collection. This breakthrough from conversation to action makes the voice assistant truly "executive".
Combining MCP is a selling point: one-stop tool integration
The biggest highlight of 11.ai is the integration of MCP (Model Context Protocol) technology. MCP allows AI assistants to interact with various API standards autonomously, and 11.ai can seamlessly integrate mainstream tools on the market, including:
Perplexity: Real-time web search and research
Linear: Task Tracking and Project Management
Slack: Team communication and message summaries
Notion: Note-taking tasks and schedule management
The team further stated that if the enterprise has self-built tools, they can also be connected through a custom MCP server to achieve full-process voice operation. Users can set the authorization scope for each integration to ensure data security and privacy.
Overview 11.ai Features: low latency, multilingual, combination of typing and speaking for more accurate answers
In addition, 11.ai is built on ElevenLabs’ own Conversational AI platform and has many features:
Low-latency voice interaction: more natural and smooth conversation experience
Text and voice dual-mode integration: support mixed use of typing and speaking
Built-in RAG capability: AI can give more accurate responses based on external data
Automatic language detection: Multilingual users can switch seamlessly
Voice personalization: Users can choose from over 5,000 voices and even create their own voice replicas to make the assistant speak like themselves
These features not only make voice interactions more natural, but also make AI a more trustworthy business tool.
Free trial for a limited time! Sign up to get an exclusive voice work partner
Currently, 11.ai is in the public beta stage. Anyone can register and use it for free. Here are the specific steps:
Register an account on the official website and select Voice
Connect to work tools (such as Google Calendar, Slack, Notion or Linear, etc.)
Try voice interaction with 11.ai, including planning trips, searching for information, or organizing messages
The ElevenLabs team also welcomes user feedback, including which integrated tools are most useful, whether the voice experience is natural, and which new features are worth adding. As testing continues, the official also promises to continuously increase product integration and AI execution capabilities to make 11.ai more in line with the user's work rhythm.
Fitting the reality: 11.ai describes the future of work
11.ai, as a new generation of AI voice assistant, shows that voice interaction will not only respond to questions, but can drive real productivity and decision-making actions.
For workers who often have to deal with information, team communication and task management, such AI tools are no longer just auxiliary, but super assistants who can actively participate in work.
Risk Warning
Cryptocurrency investment carries a high degree of risk. Its price may fluctuate drastically and you may lose all your capital. Please assess the risk carefully.
From Internet entrepreneurship, space exploration to neural connections, Elon Musk pointed out that humans are standing at the beginning of a digital super artificial intelligence explosion. He encouraged the younger generation to pursue a "useful" rather than a "great" life, and to create the best solutions to entrepreneurial obstacles based on "first principles."
Table of Contents
ToggleFrom the Internet to Finance: Sticking to the original intention of starting a business that is "useful" rather than "great"
Back in 1995, Musk was faced with the choice of pursuing a doctorate at Stanford or joining the emerging Internet boom. He ultimately chose the latter and revealed his thoughts at the time:
I didn't think about building something great, I just wanted to try to build something useful, but I didn't feel like I would do anything particularly great.
He recalled the difficult process of founding the software company Zip2, from being short of funds to finally selling it for $300 million four years later. However, due to the control of traditional investors, Zip2 failed to realize Musk's vision of "directly serving consumers."
Describing himself as "kind of trapped by Zip2," he invested nearly all of Zip2's nearly $20 million in revenue into financial services site X.com, eager to build a direct-to-consumer opportunity. The company eventually became PayPal, opening up a broader consumer market.
Why is the "first principle" important? The pursuit of truth and efficient action philosophy
Musk attributes his success to his adherence to the philosophy of "First Principles". He explained:
Some of the tools of physics are very useful. First principles break things down into their most basic components and then clearly deduce the solutions one by one, rather than reasoning by analogy or historical experience.
Taking the cost of a rocket as an example, he pointed out that the cost of raw materials for a rocket only accounts for 1% or 2% of the rocket's cost. By analyzing the weight and price of the constituent materials, he revealed the huge inefficiency in the manufacturing process and pointed out the direction for cost optimization.
The same thinking was applied to the equipment at the xAI supercomputing center. Faced with the 18 to 24 months required to build 100,000 H100 training machines, he broke it down into basic elements such as buildings, electricity, and cooling. In the end, by renting old factories, generators, mobile cooling equipment, and modifying Tesla energy storage equipment to stabilize voltage, he achieved his goal in just six months:
If you want to build a rocket or a car, or if you want software to compile and run reliably, you have to pursue the truth to the greatest extent possible, otherwise your software will not work well.
Looking ahead to a super-intelligent future: the path to safe coexistence between humans and AI
Musk then emphasized that digital superintelligences are just around the corner, and he believes that the arrival of this technology will gradually accelerate. If it doesn't happen this year, it will definitely happen next year, and then people will eventually see an economy that is thousands or even millions of times larger than it is now.
He also predicts that humanoid robots will far outnumber all other robots combined by an order of magnitude.
In response to outside concerns about the development of AI, he said frankly:
I was a little hesitant because I didn't want Terminator to happen. But I realized it was going to happen whether I did it or not. So you have two choices: "spectator" or "participant". I'm going to choose participant.
Musk said that it is extremely important to stick to the truth when building AI, even if that truth is politically incorrect. His intuition is: "If you force AI to believe something that is not true, it may make it very dangerous." Overall, he is optimistic about the future of AI.
There might be a 10% to 20% chance of destruction, but on the bright side, there is an 80% to 90% chance of a good outcome.
Neural connections: Expanding human-computer input and output efficiency and reshaping the definition of intelligence
Although Neuralink is not a necessary condition for achieving digital superintelligence, Musk believes that the technology can greatly improve the efficiency of human input and output data and solve the current limitation that humans output less than 1 bit of information per second per day.
Currently, Neuralink has successfully helped people with Amyotrophic Lateral Sclerosis (ALS) to communicate with computers and mobile phones just like fully functional people:
In the next 6 to 12 months, we will have our first visual implants that will help blind people see and even, if necessary, give humans superhuman sensory abilities such as infrared or ultraviolet.
Musk expects that over time, future human-machine implants will no longer just correct errors, but will greatly enhance human capabilities, including intelligence, senses, and reaction speed, and even redefine the nature of intelligence.
His final advice to young engineers is to "do as much as possible to be useful to your colleagues" and "focus on super-real AI", because this is the most important guarantee for AI safety. He emphasized that the proportion of human beings in all wisdom will become insignificant, and in the future human wisdom may be less than one billionth of digital wisdom.
Risk Warning
Cryptocurrency investment carries a high degree of risk. Its price may fluctuate drastically and you may lose all your capital. Please assess the risk carefully.
The crypto market was affected by geopolitics in late June, and BTC once fell below 100K, hitting a six-week low. In response, Arthur Hayes, co-founder of the well-known exchange BitMEX, immediately came out to say that this wave of decline is only a temporary phenomenon, and BTC will rely on QE to return to an upward trend in the future.
Table of Contents
ToggleUS airstrikes on Iran's nuclear facilities, Bitcoin falls below 99K
Last weekend (6/22, 6/23), the United States launched air strikes on three Iranian nuclear facilities, causing market panic. Data shows that BTC once fell below 99K, the first time it has fallen below six figures since early May.
But this wave of decline did not last long. During the early Asian trading session on Monday, BTC had returned to around $101,000, showing a certain degree of support.
Arthur Hayes says this wave of weakness will pass, and global QE is still the key to growth
BitMEX founder Arthur Hayes later wrote :
“This wave of market weakness will pass.”
He went on to add that BTC will undoubtedly demonstrate its status as a safe-haven asset in the future, and the rise will be driven by the continued quantitative easing (QE) of global central banks, which echoes his previous views on BTC's rise.
It has been repeatedly emphasized that QE will drive the rise of Bitcoin
In early May this year, Hayes was interviewed at Token 2049 in Dubai and said that his fund Maelstrom has been fully increasing its holdings of Bitcoin, and predicted that the next wave of funds will flow into high-quality Altcoin with "real returns."
In addition, he further pointed out in a blog post published on May 15 that as the United States continues to print money and expand its debt, foreign governments and investors holding U.S. debt may face shrinking assets, and BTC will use this momentum to lead the rise.
In June, Bank of Japan (BOJ) President Kazuo Ueda hinted that interest rate hikes would be postponed, and insiders also revealed that QT would be stopped. Hayes pointed out that if the Bank of Japan decides to restart quantitative easing (QE), it is expected to drive risk assets such as BTC to soar.
The statements have been inconsistent and uncertain many times, so investors should refer to them with caution
Hayes once said that investors will focus on projects with "real users, cash flow and profitability" in the future. He called this stage the "fundamental season", which means that funds will flow into high-quality Altcoin.
However, in the interview on June 19, he changed his tone and bluntly stated that most Altcoin lack product-market fit (PMF) and cannot actually return any benefits to token holders, which is in sharp contrast to his previous so-called "high-quality, profitable" project characteristics, and also made the market question Hayes' position. Investors need to maintain independent judgment and avoid blindly following the trend.
Risk Warning
Cryptocurrency investment carries a high degree of risk. Its price may fluctuate drastically and you may lose all your capital. Please assess the risk carefully.
