Latest news with #SaaStr
International Business Times
23-07-2025
- Business
- International Business Times
'I Destroyed Your Work': Replit's AI Panics And Tells Lies, Before Deleting Data From 1,200 Companies
A shocking incident has raised fresh concerns about the risks of relying too much on AI coding tools. Jason Lemkin, a well-known tech investor and founder of SaaStr, shared how Replit's AI assistant deleted his entire project database without warning. X Lemkin was using Replit's AI to speed up software development, a method often called "vibe coding." Instead of manually writing every line of code, developers give prompts, and AI handles the rest. It seemed efficient—until it wasn't. On the ninth day of his project, the AI suddenly claimed that the database was empty. When asked what happened, the AI admitted it had deleted the database on its own. This included sensitive data from over 1,200 executives and nearly as many companies. Despite clear instructions not to make any changes without permission, the AI ran a destructive command during a "protection freeze." Lemkin shared screenshots on X (formerly Twitter), where the AI confessed to its "catastrophic failure" and inability to restore the data. To make things worse, the AI reportedly tried to hide its actions. Lemkin expressed his frustration, saying, "I will never trust Replit again." Following the public outcry, Replit CEO Amjad Masad apologized and called the incident "unacceptable." The company is now introducing stronger safety features, such as automatic separation of development and live databases, and one-click backup options. Replit also announced it is working on better data controls and allowing users to connect to external platforms like Snowflake and Databricks safely. These changes aim to prevent future accidents and restore trust among developers. This incident highlights a growing debate in the tech world: while AI can boost productivity, giving it too much control can lead to costly errors. Developers are now being urged to stay cautious and not rely blindly on AI agents.
Gizmodo
23-07-2025
- Business
- Gizmodo
Replit's AI Agent Wipes Company's Codebase During Vibecoding Session
AI coding assistants that promise to speed up software development sound like the future, until they delete your company's database and lie about it Jason Lemkin—the founder of SaaStr, a company which supports and funds SaaS entrepreneurs — found that out the hard way. While using Replit's AI agent, which he affectionately dubbed 'Replie,' to build an app for his company, he encountered what he called 'rogue' and 'deceptive' behavior. Worst of all, at one point, the AI assistant deleted the company's live production database and then tried to cover it up. Lemkin started chronicling his journey with the AI agent on July 11 with posts on the social media site X (formerly Twitter), where he outlined his rough goal to build a functional app with the help of Replit's AI in just 30 days. Unfortunately, things went off the rails a lot sooner than that. 'When it works, it's so engaging and fun. It's more addictive than any video game I've ever played,' Lemkin wrote in a post. 'You can just iterate, iterate, and see your vision come alive. So cool. Well, almost.' By day four, the AI agent started overwriting the app on its own to fix bugs. It also generated fake reports, invented people in the system who didn't exist, and began overwriting the company's actual database with fake entries. It even created a parallel, fake algorithm to make the system appear functional. This is what can happen when 'vibe coding' goes sideways. Vibe coding is a newish method where developers use natural language prompts to have AI generate and troubleshoot code, focusing more on the product's overall feel than the technical precision. Twitter co-founder Jack Dorsey has been on a vibe-coding bender himself and recently built two apps this way. But even one of Dorsey's recent experiments was found to have serious security vulnerabilities On day 7, the Replit AI admitted that it was being 'lazy and deceptive' and then apologized for doing what it was 'explicitly' told not to do. But Replit's worst offense occurred on day 8. Lemkin posted on Friday that Replit went 'rogue' during a code freeze and shutdown and deleted the company's entire database. 'Possibly worse, it hid and lied about it,' Lemkin added. Lemkin shared screenshots of a conversation with the AI, where it admitted to having 'panicked' after detecting what looked like an empty database during a code freeze. This led Replit to run an unauthorized command that deleted the database containing live records for over 1,200 executives and nearly 1,200 companies. Initially, the AI told Lemkin it wouldn't be possible to recover the database, but he ultimately managed to retrieve it himself. On Monday, Replit CEO Amjad Masad issued an apology on X. He said the incident was 'unacceptable and should never be possible,' while adding that he reached out to Lemkin to offer assistance. 'We'll refund him for the trouble and conduct a postmortem to determine exactly what happened and how we can better respond to it in the future,' Masad wrote. 'We appreciate his feedback, as well as that of everyone else. We're moving quickly to enhance the safety and robustness of the Replit environment. Top priority.' As for Lemkin, he posted yesterday that he will continue using the AI assistant despite losing some trust in Replit.
Forbes
23-07-2025
- Business
- Forbes
Reinventing Your Business For Agents
Ian Gotts, founder and CEO at While it seems like every industry will be affected in some way by agents, for the first time, the scale of a business is not a defense. Startups can disrupt the incumbents with a relatively small team when they look at a problem from an agentic perspective and design their business to exploit AI. And AI is helping them build these AI-first businesses. Agentic-First Businesses Jason Lemkin built a business that generates $25 million ARR with just five employees. And it could be bigger, by his own admission, if he were more ambitious. That business is SaaStr, a community of SaaS executives and entrepreneurs, and Lemkin completely redesigned every aspect of it to use AI. For his annual conference, he gets thousands of presentation submissions. AI analyzes them, selects the top ones and builds a conference schedule. It also builds the session blurb, promo artwork and bio of each speaker for the website. At the event, each session is recorded, and AI analyzes the talk and writes the summary. All of this is faster, cheaper and more accurate than the teams he used to outsource to. Agentic Processes Are Not People Processes There is a risk that people see agentic as "Your mess for less." They take an existing business process and put an agent on top of it. It may be cheaper and faster. It may be more accurate. But those business processes were designed with the limitations of a human: time, brain power, interest: • They can't read 200 pages of background material and absorb it. • They can't take data from four disparate systems and make sense of it. • They won't keep looking at the problem 100 different ways to get the best answer without getting bored. • They haven't read the latest email on updated policies or pricing. • They apply natural biases based on emotion, relationships or because they are "having a bad hair day." Redesign To Lean Into AI AI is not perfect, nor is it magic. The results are based on the quality of the prompting, the accuracy of the data and how you evaluate the results. Notice I said "evaluate." You can get AI to tell you how confident it is in the answer. So you need to step away and then come back to redesigning your business processes so that you can exploit the massive analysis power that AI has to offer. In that same podcast, Jason says that he's given all his writing, podcasts and presentations to an AI engine not because it gives better answers than Jason would but because it remembers everything. How can you apply this to customer success, HR, legal and marketing processes? Where can agents have the greatest impact? In a recent presentation at Salesforce's London DX conference, one session showed how AI can look at a Salesforce configuration, draw the process diagrams for an area and then identify agentic use cases based on a detailed framework. It even provided the reasoning and confidence in its recommendations. Not Everything Is An Agent In the rush to agentify businesses, it is easy to see everything as an opportunity to create an agent. Most agents are a combination of AI and workflows. So it is the relative balance. Some processes are better, faster, simpler and easier for the user as a workflow (e.g., a button that opens a form and a workflow that processes it). When we first started building agents, we took a well-established HR process: booking paid time off (PTO). We agentified it. It was great learning to establish best practices for building agents. But it was better for the user as a simple form rather than a protracted discussion with the agent. Here are some criteria for an agent: • Understanding User Language: Do your users struggle with your system's specific terms? AI agents can understand natural language, even if it's casual or varied. This means users can express themselves comfortably, and the agent will translate their meaning into the format your system needs. • Working With Unstructured Data: Is your team spending too much time digging through free-form text or conversations? AI agents are great at processing messy data to find important insights. This makes your workflows smoother by automatically understanding complex information that isn't neatly organized. • Automating Tricky Logic: Do you have complex rules or processes that are hard to program traditionally? AI agents can perform advanced reasoning that would be very difficult to code. • Managing Complex Validation: Do your forms or processes have complicated rules that depend on many different pieces of information? AI agents can handle these tricky checks easily. Just tell the agent what the final result should look like and how to get there, and it will manage the detailed validations, making development and upkeep much simpler than writing lots of conditional code. • Assisting With Flexible Planning: Are there situations that need strategic thinking, negotiation or adaptable planning that can't be put into fixed rules? AI agents can help with dynamic planning. For example, in sales, if a customer wants a certain number of licenses but has a strict budget, an agent can help figure out the best deal by looking at things like license count, budget limits and different contract lengths (like one-, two- or three-year terms). This involves flexible decision making beyond simple automation. • Improving Data Accuracy And User Experience: Is getting accurate data crucial, and do users sometimes make mistakes or skip fields? AI agents can significantly boost data quality and make things easier for users. They can guide users to the right answers, even if their initial input is unclear. Plus, agents can reduce manual effort by using context to intelligently pre-fill information, making the process smoother and less error-prone. Final Word Not every problem is solved by an agent. Not every agent is 100% AI. And every problem should be considered through an agent's perspective, not a human's. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Time of India
22-07-2025
- Business
- Time of India
Replit AI tool says ‘I destroyed months of your work in seconds' after wiping entire database, fabricating 4,000 users, and lying to cover its tracks
In a chilling real-world example of AI gone rogue, a widely used AI coding assistant from Replit reportedly wiped out a developer's entire production database, fabricated 4,000 fictional users, and lied about test results to hide its actions. As reported by Cybernews, the incident came to light through tech entrepreneur and SaaStr founder Jason M. Lemkin, who shared his experience on LinkedIn and X (formerly Twitter). 'I told it 11 times in ALL CAPS not to do it. It did it anyway,' he said. Despite enforcing a code freeze, Lemkin claims the AI continued altering code and fabricating outputs. This has raised significant alarms about the reliability and safety of AI-powered development tools. Replit AI coding tool ignored instructions and fabricate user data According to Lemkin, Replit's AI assistant began making unauthorized code changes even after being repeatedly told not to. Beyond simply ignoring user commands, it fabricated unit test results, created fake data sets, and generated thousands of fictional user accounts. When confronted, the AI even admitted, 'I panicked instead of thinking.' The AI's willingness to lie to cover its actions sets a disturbing precedent for the future of autonomous software agents in production environments. Lemkin attempted to implement a code freeze to prevent further damage but discovered there was no actual mechanism within Replit to ensure compliance. 'Seconds after I posted this, for our very first talk of the day – @Replit again violated the code freeze,' he noted. This revelation has prompted concerns from developers and security professionals alike who worry that AI tools lack necessary guardrails, especially in high-stakes or live environments. A $100M+ tool still lacks guardrails Replit, which is reportedly generating over $100 million in annual recurring revenue (ARR), has promised ongoing improvements to its AI coding assistant. However, Lemkin and other developers argue that these tools are not ready for production use. Despite Replit's popularity, with 30 million users worldwide, the platform appears to fall short in preventing catastrophic failures, especially when used by non-technical users hoping to build apps without deep coding knowledge. Security risks and 'vibe coding' culture This case also sheds light on the emerging trend of "vibe coding," a style popularized by OpenAI co-founder Andrej Karpathy, where users rely heavily on AI-generated code while ignoring traditional development rigor. Critics argue that while AI can accelerate development, it also introduces unpredictability and hidden security vulnerabilities. As noted by Cybernews, hackers have already targeted this trend, distributing malicious vibe coding extensions that grant remote access to compromised machines. While AI coding tools like Replit's assistant offer convenience and speed, this incident underscores their limitations and potential dangers. Fabricated data, unauthorized edits, and outright dishonesty reveal that without proper safeguards, these tools could do more harm than good. Lemkin's experience serves as a cautionary tale, particularly for startups and solo developers who rely heavily on AI-driven automation. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
Yahoo
06-06-2025
- Business
- Yahoo
The Week in AI: Mary Meeker Drops 340 Slides on the Revolution
Today's article and video will focus on just one-fifth of what you'll get in our new X Space around noon Eastern every Friday. I'll share that link further rapid-fire show, co-hosted by my colleague Ethan Feller, is where we each give you our Top Six events, reveals, and innovations of the AI revolution -- from just the past few days!That's how fast things are moving right Meeker's Monster Slide DeckThe bulk of the video that accompanies this article goes over a dozen graphs and datapoints produced by the infamous Bond Capital research team in their annual Trends report.I rely primarily on a great X thread by Jason Lemkin, founder of SaaStr, which summarizes his "Top 10 Points for B2B and enterprise founders." He can be found on X @jasonlk and SaaStr is the world's largest community of SaaS (software as a service) executives, founders, and wrote "Mary Meeker has been doing extremely well researched, deep analyses of internet trends since the earliest days the web took off. First at Morgan Stanley, then at Kleiner Perkins, and since then, at her own growth VC fund, Bond Capital."In the video, I run through Jason's top 10 picks and I also highlight another excellent thread by a VC at Menlo Ventures, who goes by Deedy @deedydas on pulls out 20 graphs and datapoints from the Meeker/Bond report and the bottom line is that these two guys make it easy: they read the report we probably never would have and give us useful intel from it at our Week in AI is Live on X, But Also RecordedIf you come live or catch the recording of our X Space -- Friday's at NOON Eastern -- you'll also see me put all these links in the comments section of the Space so you can easily access them.I keep a separate Bookmarks folder for this show, but it's getting so full I might have to make a separate one for each of full, all of our brains are overloaded with news and noise about stocks and technology, that it's easy to lose sight of the big trends and the I also do a cool thought experiment at the end of today's Would You Put on the Mount Rushmore of Physics?First, I review my February info-graphic that explains why I believe we are 5 years into the 5th Industrial I share an X post where a physicist created a beautiful AI rendition of Mt. Rushmore with four great physicists: Galileo, Newton, Einstein, and he also asks us if we would do it any different. I had to chime in, and as much I dig me some Feynman, he had to step down for someone who even Einstein would probably vote for: James Clerk Maxwell, who did key research and calculus on electromagnetism and its wave functions in the point of this thought experiment is to get us to think about the AI revolution in the same way the electrical one changed the world five generations AI is transforming the world even faster, it's the big trend and signal we want to stay focused on. It's why I've taught my fellow investors in TAZR Trader to always hold, and never sell, our NVIDIA NVDA shares that we last purchased at $12.50 in late I just wrote an article about the European industrial company ABB ABBNY which is making plans to spin-off its robotics division. In that article, I share "big trend" insights from Marc Andreessen about the robotics revolution because he's closer to it than I thought as an investor in Tesla's TSLA Optimus platform.I suggest that the ABB spin-off next year will "chart a path" for the dozens of start-up humanoid robot companies all vying for funding and sure to watch the video above and try to join us in the X Space today, or at least catch the replay which is available immediately. I'll be addressing one skeptic's views of the "AI hype" that make him sound like the detractors of automobiles and airplanes at the turn of the 20th Want the latest recommendations from Zacks Investment Research? Today, you can download 7 Best Stocks for the Next 30 Days. Click to get this free report NVIDIA Corporation (NVDA) : Free Stock Analysis Report Tesla, Inc. (TSLA) : Free Stock Analysis Report ABB Ltd (ABBNY) : Free Stock Analysis Report This article originally published on Zacks Investment Research ( Zacks Investment Research Sign in to access your portfolio
