IP.com

IP.com IP.com is driving innovation forward, faster. IP.com was founded with an entrepreneurial spirit and a relentless desire to make the complex simple.

Our suite of powerful intellectual property solutions, including innovation analysis tools and patent search services, makes complex data actionable. We are driven to deliver the most intuitive and powerful intellectual property tools that enable organizations to make critical business decisions. We believe time is money and information is power. And we work relentlessly to deliver both for our customers. Our powerful suite of analytic tools provides insights that drive innovation.

The Supreme Court Just Let an Abandoned Patent Application Kill Real IP. Here’s What That Means for Patent Holders: At t...
06/01/2026

The Supreme Court Just Let an Abandoned Patent Application Kill Real IP. Here’s What That Means for Patent Holders: At the center of this case is a gap in U.S. patent law that doesn’t get much attention outside of litigation circles, but that every patent holder should know about: the enablement standard for getting a patent granted is meaningfully different from the enablement standard for prior art used to challenge one. Read our blog here: https://buff.ly/9mO1Nt6

Who Is the Skilled Person Now? How AI Is Rewriting Patent ObviousnessFor IP counsel and patent professionals, this is a ...
05/28/2026

Who Is the Skilled Person Now? How AI Is Rewriting Patent Obviousness

For IP counsel and patent professionals, this is a debate worth following closely. The resolution, whenever it comes, will affect how claims are drafted, how portfolios are valued, and how validity is argued in litigation. Read our full blog here: https://buff.ly/j17Lt3l

The AI That Sees Cancer Three Years Before Doctors Can: Pancreatic cancer is one of medicine's cruelest diagnoses — not ...
05/27/2026

The AI That Sees Cancer Three Years Before Doctors Can: Pancreatic cancer is one of medicine's cruelest diagnoses — not because it can't be treated, but because it's almost never caught in time. A new AI model from Mayo Clinic may be about to change that.

REDMOD doesn't require new equipment, new procedures, or special scans. It runs on the routine CT imaging patients already receive — and in it, the AI finds something human eyes simply cannot: invisible tissue patterns, textural shifts, and structural anomalies that quietly signal a cancer forming years before any radiologist would flag a concern.

The results from nearly 2,000 CT scans — all originally read as normal by specialists — are striking. REDMOD correctly identified 73% of early pancreatic cancers that had gone undetected. At the two-year mark before diagnosis, the gap between AI and human performance became dramatic: the model spotted roughly three times as many early-stage cancers as experienced radiologists did. Not marginally better. Three times.

What's powering this is the kind of pattern recognition AI uniquely excels at — processing hundreds of quantitative imaging features simultaneously, detecting subtle correlations across texture and structure that fall entirely outside the range of human perception. This isn't AI assisting a radiologist. It's AI seeing things a radiologist fundamentally cannot.

The stakes couldn't be higher. Pancreatic cancer's five-year survival rate sits below 15% — a number that has barely budged in decades, largely because by the time symptoms appear, the disease has already won. Earlier detection doesn't just improve odds; it transforms them.

What makes REDMOD genuinely exciting is the path to scale. Because it works on standard CT scans rather than specialized imaging, early AI screening doesn't have to be a separate, friction-heavy diagnostic step — it can simply become part of what already happens. Quietly. Automatically. Three years earlier than before. Read more here: https://buff.ly/QarEvrX

AI Just Learned to Taste — Without Ever Eating. No chemistry data. No flavor labels. No sensory panels. Just recipes — a...
05/21/2026

AI Just Learned to Taste — Without Ever Eating. No chemistry data. No flavor labels. No sensory panels. Just recipes — and the patterns buried inside them. That's the premise behind Epicure, a new paper from food robotics startup KAIKAKU AI, and the results are hard to dismiss. Trained only on how chefs combine ingredients, the model independently identified all five basic tastes, correctly ranked peppers by spiciness, and sorted cuisines by geographic region. It learned the grammar of flavor the same way a language model learns the grammar of prose — not by being told the rules, but by inferring them from structure.

The underlying dataset is modest but carefully constructed: 6,653 raw ingredient entries cleaned down to 1,032 usable foods, then mapped across recipes to reveal how ingredients cluster, contrast, and co-occur. From that alone, the model built what amounts to a working theory of taste — one that aligns with human sensory science without ever having been exposed to it.
KAIKAKU is positioning this as a "ChatGPT moment" for food AI, and the analogy has genuine merit. Just as large language models revealed that statistical patterns in text encode meaning, Epicure suggests that statistical patterns in recipes encode something real about flavor, texture, and culinary logic. The three applications the team identifies — menu development, recipe innovation, and flavor pairing — are all domains currently governed by chef intuition and institutional knowledge that is notoriously difficult to systematize.

The longer play is more ambitious. KAIKAKU intends to pair this model with its robotics platform, pitching the combination as "autonomous food infrastructure" for commercial kitchens — an operation that doesn't just execute recipes but understands them.
Recipes are, at their core, a compressed record of human preference: every combination, substitution, and proportion a signal about what works. If AI can read that signal reliably, the implications extend well beyond menu suggestions — into product development, dietary substitution, and the design of food experiences built around learned taste rather than educated guesswork. https://buff.ly/fQeArEF

Bissell v. ITC: Firmware Evasion and the Limits of a Limited Exclusion Order:A successful Section 337 investigation does...
05/20/2026

Bissell v. ITC: Firmware Evasion and the Limits of a Limited Exclusion Order:

A successful Section 337 investigation does not necessarily yield durable relief. A sophisticated respondent will redesign — and in an era where product functionality resides in software, redesign often means a firmware update. The Federal Circuit's decision in Bissell, Inc. v. International Trade Commission, Nos. 2024-1509 & 2024-1709 (Fed. Cir. May 11, 2026), illustrates both the ease with which that evasion can be accomplished and the doctrinal constraints that may leave a prevailing patent holder without an effective remedy.

Background:

Bissell's U.S. Patent No. 11,076,735 claims a wet/dry floor cleaner with an automatic self-cleaning cycle containing a negative limitation: the battery charging circuit "is disabled by the actuation of the self-cleaning mode input control and remains disabled during the unattended automatic cleanout cycle." Following Bissell's March 2022 complaint, respondent Tineco deployed a firmware update altering the 120-second cleanout cycle so that the charging circuit activates briefly twice during the cycle rather than remaining off throughout. The modification served no independent engineering purpose; its apparent object was to defeat literal infringement.

Holdings and Implications

The ITC found the original products infringing and entered a limited exclusion order, but declined to find infringement as to the redesigned firmware variants, rejecting Bissell's doctrine of equivalents argument. The Federal Circuit affirmed in full. The presence of an express negative limitation appears to have substantially constrained the equivalents analysis — a result that, on these facts, rewards an insubstantial workaround of transparent litigation motivation.

The decision crystallizes a structural vulnerability in the limited exclusion order regime: because such orders run against specifically identified products rather than a respondent's downstream output generally, a timely firmware update can render a hard-won exclusion order prospectively toothless — with doctrine of equivalents providing no reliable safety net.

AI Gets Its Biggest Shot at Understanding Life ItselfWhat if an AI could predict exactly how a human cell behaves — or m...
05/19/2026

AI Gets Its Biggest Shot at Understanding Life Itself
What if an AI could predict exactly how a human cell behaves — or misbehaves? That's the animating question behind a landmark $500M bet on AI-driven biology. CZI Biohub, the nonprofit research organization backed by Mark Zuckerberg and Priscilla Chan, has launched the Virtual Biology Initiative — a half-billion-dollar push to build the foundational data and models needed to teach AI how living systems actually work. The ambition isn't incremental. It's to simulate biology the way physicists simulate particle collisions: computationally, at scale, and with predictive power.

The bulk of the funding — $400M — goes directly toward data generation and advanced imaging technology, with the remaining $100M flowing to external research labs. The coalition forming around it is notable: Nvidia, the Allen Institute, Arc, and others are signing on, with Biohub committing to open datasets that the broader AI research community can build on.

The data gap is the crux of the challenge. Today's best AI biology datasets top out around 1 billion cells — impressive until you consider that Biohub's own Alex Rives argues the field needs an order of magnitude more to meaningfully accelerate progress. That's not a rounding error. It's a fundamental constraint on what current models can learn.
The endgame? Training AI systems capable of "understanding disease and reprogramming it at the level of cells, molecules, and tissues" — language that echoes Demis Hassabis's long-held conviction that AI could one day effectively end disease as we know it.

The open question is whether the same scaling dynamics that cracked language understanding and protein folding will transfer to the staggering complexity of cellular biology. $500M is a serious commitment — but against the data requirements Rives describes, it may be just the opening move. Red more here: https://buff.ly/O3elTSp

Joyce Hill and the IP.com team are live at the FLC National Meeting, giving demos of our innovation and IP intelligence ...
05/14/2026

Joyce Hill and the IP.com team are live at the FLC National Meeting, giving demos of our innovation and IP intelligence solutions. Stop by our booth to see the solutions in action and learn how we can help you and your team move innovation forward faster. Plus, schedule a demo for your chance to win an iPad. Check the FLC app for yesterday’s winner announcement.

Meta Just Built a Brain You Can Run on a Laptop:For decades, understanding the human brain meant one thing: putting peop...
05/14/2026

Meta Just Built a Brain You Can Run on a Laptop:
For decades, understanding the human brain meant one thing: putting people inside a $3 million scanner and hoping they didn't move. That bottleneck may now be over.
Meta just open-sourced TRIBE v2 — an AI model trained on over 1,000 hours of real brain data from 700+ people that can simulate neural activity across vision, hearing, and language. No scanner required. And in a result that should genuinely unsettle neuroscientists: its synthetic predictions are outperforming real fMRI recordings.

Let that sink in. A model is now a more reliable picture of how populations of brains respond to stimuli than the actual scans themselves — because real fMRI data is messy, contaminated by heartbeats, micro-movements, and biological noise that no amount of careful methodology can fully eliminate. TRIBE v2 doesn't have that problem.

The leap from v1 to v2 is staggering: from 4 subjects to 700+, from 1,000 brain regions to 70,000, and from a research curiosity to something that can replicate decades of neuroscience findings in software — correctly identifying the brain regions responsible for faces, speech, and text without running a single scan.

Think about what that means. Experiments that once required IRB approvals, MRI time slots, and months of data collection can now be run in seconds on commodity hardware. Meta released the code, model weights, and a live demo — meaning any researcher on Earth can start running virtual brain experiments today.
https://buff.ly/Sl28fSF

If you’re at the FLC National Meeting in Seattle, come visit IP.com at Booth 12.See how we help innovation and IP teams ...
05/13/2026

If you’re at the FLC National Meeting in Seattle, come visit IP.com at Booth 12.

See how we help innovation and IP teams move faster with better intelligence, and book a demo for a chance to win an iPad.

See you there.

Google has officially brought Gemini to the Mac desktop, and the AI wars just got a whole lot more interesting. A year a...
05/13/2026

Google has officially brought Gemini to the Mac desktop, and the AI wars just got a whole lot more interesting. A year after ChatGPT and Claude planted their flags on the platform, the world's most powerful distribution machine has entered the chat — and the desktop AI race is now fully on.

The app snaps open with Option+Space and immediately feels like it belongs there. Screen sharing, Google Drive and Photos integration, AI image generation through Nano Banana, and video creation via Veo — it's a creative and productivity powerhouse sitting one keystroke away. At a moment when people are using AI for everything from drafting emails to running experimental research to debugging code, having that kind of capability baked into your desktop is a genuinely big deal.

Yes, Gemini is still catching up on the agentic side — Claude and ChatGPT can already take the wheel and execute tasks directly on your machine, while Gemini is currently playing the role of an exceptionally smart assistant rather than an autonomous operator. But Google has called this release "just the beginning," and if there's one company on the planet with the resources and reach to close that gap fast, it's Google.

A Windows app landed alongside it too, fusing Gemini and Google Lens into a sleek unified search bar. Global languages are coming — but the momentum is undeniable.
We're living through a moment where AI is becoming the operating layer on top of everything — our phones, our browsers, our workflows, our finances. The desktop is just the latest frontier, and now every major AI player has a seat at the table. Gemini showing up late matters far less than the fact that it showed up — and Google's billion-plus users mean this could scale faster than anything we've seen yet. https://buff.ly/jwJJ7l0

Address

370 Woodcliff Drive Suite 301
Fairport, NY
14450

Opening Hours

Monday 8am - 8pm
Tuesday 8am - 8pm
Wednesday 8am - 8pm
Thursday 8am - 8pm
Friday 8am - 8pm

Telephone

+18664736826

Website

Alerts

Be the first to know and let us send you an email when IP.com posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to IP.com:

Shortcuts

Share

Category