InfoCommerce Group

🦔A researcher invented a fake eye condition called bixonimania, uploaded two obviously fraudulent papers about it to an academic server, and watched major AI systems present it as real medicine within weeks. The fake papers thanked Starfleet Academy, cited funding from the Professor Sideshow Bob Foundation and the University of Fellowship of the Ring, and stated mid-paper that the entire thing was made up. Google's Gemini told users it was caused by blue light. Perplexity cited its prevalence at one in 90,000 people. ChatGPT advised users whether their symptoms matched. The fake research was then cited in a peer-reviewed journal that only retracted it after Nature contacted the publisher. My Take The researcher made the papers as obviously fake as possible on purpose. The AI systems didn't catch it. Neither did the human researchers who cited it in real journals, which means people are feeding AI-generated references into their work without reading what they're actually citing. I've covered the FDA using AI for drug review, the NYC hospital CEO ready to replace radiologists, and ChatGPT Health launching this year. All of that is happening in the same environment where a condition funded by a Simpsons character and endorsed by the crew of the Enterprise was being presented as emerging medical consensus. The people making these deployment decisions seem to believe the pipeline from research to AI to patient is more supervised than it actually is. This experiment suggests it isn't supervised much at all. Hedgie🤗 nature.com/articles/d4158…

A friend of mine from Harvard Law set up his own firm last year. Solo practice. No associates. No paralegals. Working out of a co-working space with a laptop and a coffee habit. Last month a mid-size business owner reached out looking for outside counsel. Three firms were being considered. Two of them were 15-attorney shops. The kind with pitch decks, associate teams, and glass-walled conference rooms that smell like fresh carpet and overbilling. My friend was a one-person firm with a WeWork membership. He almost canceled. He thought there was no way he could compete with that. I told him to try one thing before he walked away. Open Claude Code. Give it the owner's name, the company name, and 45 minutes. Ask it to build a complete intelligence report using only publicly available data. He did not think it would work. He tried it anyway. Claude came back with a 13-page report. He read it over coffee. Took 28 minutes. By the time the Zoom started, this solo attorney knew things about the prospect's company that the owner's own in-house team probably had not assembled in one place. The company was incorporated in Delaware but registered as a foreign entity in Texas 14 months later. That is expansion. A second member was added to the LLC in 2024. Claude pulled the operating agreement implications from the state filing and flagged what a new member meant for governance, profit distribution, and decision-making authority. Three active trademark applications filed in the last six months. Two were in a product category the company had never publicly announced. Nobody on the website knew about it. The trademark filings did. PACER hit. The company had been named as a defendant in a vendor dispute 18 months ago. It settled. But the complaint was public and Claude read every page of it. The core issue was a supply agreement with no termination clause. My friend now knew this company had been burned by a bad contract. They would care deeply about airtight vendor agreements going forward. He did not have to guess. It was in the filing. State court records. The owner had a dissolved LLC from 2019 with a different partner. A business divorce. Which meant this owner would value clear partnership terms and buy-sell provisions this time around. People who have been through a bad breakup want a prenup for the next one. Same principle. Hiring activity. Four job listings posted in 60 days. Head of compliance. Operations manager. Two warehouse roles. They were scaling fast and hiring operational infrastructure. That is exactly when companies need outside counsel the most and know it the least. They think they need a lawyer when they get sued. They actually need a lawyer when they start hiring a Head of Compliance. Glassdoor. 11 reviews. Every positive one mentioned culture. Every negative one mentioned the same thing. "No HR. No handbook. No process." A company growing faster than its internal policies. An employment claim waiting to happen. And a business owner who probably had no idea what his own employees were writing about him. Google reviews. 4.3 stars. But Claude flagged a pattern in the 1-stars. Three different customers mentioned the same issue. Product delivered late with no communication. The biggest operational liability was not product quality. It was fulfillment. That is a breach of warranty problem, a customer retention problem, and a potential class issue if the pattern scales with the company. Then there was a section Claude titled "Founder Mindset." It pulled a transcript from a podcast the owner appeared on and analyzed his communication patterns. One quote stood out. He said "I have spent more on lawyers fixing problems than I ever spent on lawyers preventing them." That one sentence told my friend exactly how to position his entire practice. Not as a litigator. Not as a fixer. As the lawyer who prevents the problems in the first place. The pitch wrote itself. Claude also analyzed the owner's communication style across LinkedIn posts, podcast answers, and X replies. Based on patterns it flagged what mattered for the meeting: this person values substance over rapport. He distrusts anything that feels like a pitch. Lead with what you know. Skip the small talk. Show your work before you ask for the engagement. My friend adjusted his entire approach based on that analysis. The Zoom started. No pleasantries. No "let me tell you about my firm" warmup. The owner gave his overview. What the company does. Where they are heading. What they need. Then my friend said "I noticed you filed two trademarks in a new product category last quarter. Is that the line you are launching in Q3?" Silence. "How do you know about that?" A solo lawyer working from a coworking space just earned more credibility in one sentence than the 15-attorney firm earned in their entire pitch deck. He walked the owner through everything. The vendor dispute and what it meant for future contracts. The hiring pattern and the compliance risk it signaled. The Glassdoor reviews pointing to an HR exposure. The fulfillment complaints that were one bad quarter away from becoming a warranty liability. He did not pitch his services. He showed the owner his own blind spots using the owner's own public data. Then he said which ones he would fix first and why. The owner said "the other firms sent me a brochure. You just showed me you already understand my business better than they do." He hired my friend that week. A solo practitioner over two 15-attorney firms. No associate team. No paralegal pulling research. No marketing department. One Harvard Law grad with Claude Code, a 13-page report, and 28 minutes of preparation that the other firms did not think to do. This is what I keep telling solo lawyers and most of them do not believe me until they see it. The advantage is not firm size. It is not headcount. It is not a fancy office or a partner track or a receptionist who offers sparkling water. The advantage is showing up knowing things the prospect did not expect you to know. That is what wins the engagement. Every time. And right now it is easier than it has ever been. Because almost everything about a business is public. It is just scattered across 15 different sources that no lawyer checks before a pitch meeting. Claude checks all of them in one run and hands you a report you can read before your coffee gets cold. Secretary of State filings. Incorporation, officers, registered agents, foreign qualifications. PACER and state court dockets. Every lawsuit, motion, and settlement. USPTO. Trademark filings tell you where a company is going before they announce it. LinkedIn job postings. What a company is hiring for reveals what is broken inside. Glassdoor. What employees say when nobody from management is reading. Google reviews. The 1-star reviews are where the legal risks hide. Podcast transcripts. The founder's own words analyzed for how they think and decide. UCC filings. Who they owe money to. What assets are pledged. Property records. Leases, liens, ownership structures. Communication pattern analysis. How this specific person talks, processes information, and makes decisions. So you know exactly how to show up. All public. All free. One report. Under 30 minutes to read. The solo lawyer who builds this into their pre-meeting workflow will win clients over firms 10 times their size. Not once. Every time. Because nobody expects a solo to show up that prepared. And that gap between what they expect and what you deliver is the most valuable asset in your practice. My friend is a Harvard Law grad. He has no team. He works from a coworking space. He is winning over 15-attorney firms because he spends 45 minutes doing what they never bother to do. The playing field was never about resources. It was about preparation. And preparation just got automated.

Larry Ellison $ORCL highlighted something critical: models like ChatGPT, Gemini, Grok, and Llama are all trained on largely the same public internet data. When everyone trains on the same information, models inevitably converge. That’s why AI is moving toward commoditization. The real moat isn’t the model itself. It’s the proprietary data behind it. Companies that can train on exclusive datasets gain an advantage competitors can’t replicate. Having data that no one else has will allow you to dominate your market.

Watch Gemini 3 Flash transform manual ETL into a near real-time workflow. This demo shows how the model’s speed and reasoning capabilities instantly resolve mismatched sources, clean messy data, and fully automate the pipeline.

LLMs/general agents still struggle to make sense of messy and complex Excel data. You can't easily dump all cells into the context window, and using the code interpreter is inefficient. LlamaSheets is one of my favorite releases from last year. We've embarked on an effort to build state-of-the-art algorithms and models to segment and parse complex Excel tables - including merged cells, hierarchical rows/columns. This includes both sheet-level and table-level understanding. We think there's a ton of use cases that this can help solve (simplest example: structuring your income/P&L/cash statements to be LLM-ready), and we'd love to get your feedback. Come check it out and let us know your thoughts! Sign up: cloud.llamaindex.ai Docs: developers.llamaindex.ai/python/cloud/l…

So dang enchanted with the Gemini 3.0 Flash support in @Browserbase, especially for tedious data tasks! Check out the video below - Gemini navigates to the @USGS website for seismic data, filters on 3D Multichannel Seismic datasets, adds all of the options for California, and then downloads as a CSV list. ❤️ You could even then have the data science agent in @GoogleColab post-process and visualize the surveys!

Finds a person's profile across 1000+ websites

ByteDance just dropped an OCR model that reads documents just like humans. This 0.3B model analyzes page layout first, then parses elements in parallel. 100% open-source.

You can now snap a picture of your data and turn it into a spreadsheet. Watch this Boost Bite to see how Gemini instantly converts any image into a fully functional Google Sheet. → goo.gle/4nhE9la

As long as this is true, ChatGPT and Google will never be able to compete with Grok on a long enough timeline. Water seeks its own level.

8️⃣ Use Flash Fill to fill data at once:

No AI should be this fast at scraping. 2 clicks and it auto-crawls, scrapes entire sites... all subpages, tables, PDFs, images, docs, even hidden content & structures it perfectly. Interns, freelancers and old tools can go home. Tutorial + use cases: 🧵

i just watched a guy scrape google maps with an ai agent and show EXACTLY which boring businesses can make you stupid money in your local area not HVAC or plumbing...the local startup ideas NOBODY talks about how it works (save this) 1/ a workflow that finds overlooked local niches like garage organizers and irrigation systems through google map's treasure chest of data 2/ an agent that measures demand by tracking review volume and velocity in google maps 3/ a system that turns customer complaints into newsletter content automatically 4/ a workflow hosted on n8n for $7/month that runs nonstop and gets you concurrent workflows so dont hit a ceiling on workflows 5/ a directory that sells leads back to service providers for $100–200 each 6/ a process that scales to $20–30k/month with minimal overhead @boringmarketer shared every step on @startupideaspod and i thank him for sharing that sauce FOR FREE (people charge $10k+ for this sorta stuff) cool thing is he started a mobile diesel company in north carolina using this exact process that's making $30k/month so he knows it works what hit me is how obvious it felt. of course google maps has the gold just need AI to shovel it out.

To my knowledge, this is the first publicly available precinct-level estimate of Presidential support by demographic on a national scale. Here’s a quick demo of how it works:

We are deeply thankful to the innovative software companies that trusted us with their most important assets: their proprietary data and their data supply chains. bit.ly/4g3m4Fr

web scraping and browser automation with Crawlee

I just love this. The new =COPILOT() function in Excel lets you analyze, generate content, and brainstorm directly in the grid.

This is the reason why Renaissance hires PhDs to clean data 😬

WTF... This scraper works like a self-driving car. You give it a destination… it plans the route, clicks through pages, cleans it up & delivers a perfect spreadsheet. Goodbye manual scraping, Excel & data teams. Tutorial + some crazy examples: 🧵