In what has become a quarterly necessity for the tech industry, TechOnion presents our updated encyclopedia of Google’s AI products, initiatives, moonshots, fever dreams, and “definitely still supported” projects that have already been secretly marked for execution. This living document remains the only comprehensive record of Google’s increasingly fractal AI ecosystem, which now features more distinct products than there are engineers at the company to maintain them.
The Naming Conventions, or “Did an Algorithm Generate These Product Names?”
Google’s AI product naming strategy appears to involve throwing darts at a board containing Greek mythology references, random nouns that tested well with focus groups, and the names of the VP of Product’s childhood pets. Current naming trends include:
The “Just One Word That Sounds Vaguely Scientific” category (Bard, Gemini, Pathways)
The “Literally Just Letters” approach (LaMDA, PaLM, MUM, TPU)
The “Optimistic Verbs That Imply Capability Rather Than Describing Function” method (Discover, Lens, Search, Assist)
The “We Actually Spent $2 Million on This Branding” collection (Vertex AI, Duet AI, Med-PaLM)
According to former Google naming strategist Jennifer Nomenclature, “The key is to pick something that sounds impressive but meaningless enough that when the product pivots three times before launch, you don’t have to change the name.”
The Actual Products (As of This Morning, Subject to Change by Lunchtime)
Gemini – Formerly Bard, formerly almost called “Apprentice” until someone who’d seen The Apprentice TV show raised their hand in a meeting. Google’s flagship conversational AI that combines the knowledge-seeking capabilities of Google Search with the hallucination capabilities of that friend who always claims they “read an article somewhere on Reddit” about whatever topic you’re discussing. Currently available in three sizes: Nano (provides wrong answers quickly), Pro (provides wrong answers with greater eloquence), and Ultra (provides wrong answers that sound so authoritative you question your own knowledge).
Bard – Officially retired as a product name but still appears on approximately 73% of Google’s documentation, marketing materials, and employee email signatures due to what internal sources describe as “extreme naming inertia.” When users ask what happened to Bard, Google representatives are instructed to look pensively into the distance and say, “Bard lives on in all of us, in a way.”
Google Assistant – The AI assistant that lives in your phone, speaker, refrigerator, and possibly your dreams. Originally launched as a competitor to Siri and Alexa, it has since evolved into a complex ecosystem of its own, with seventeen different activation systems, nine separate backends, and the unique ability to understand everything you say perfectly until you actually need it to perform a useful function. Currently engaged in an internal civil war with Gemini for control of Google’s assistant ecosystem, like two Roman generals fighting for the empire while barbarians (Apple and Amazon) wait at the gates.
AIFI (Artificial Intelligence for Intelligence) – A meta-AI platform announced in 2024 designed exclusively to help Google employees understand which AI product they should be using or working on at any given time. The system reportedly crashed during initial testing when asked to map dependencies between AI projects, creating what engineers described as “the world’s most expensive stack overflow.”
Anthropic Claude Integration – Google’s $2 billion investment in Anthropic, described by one anonymous executive as “our backup plan if our own AI models decide they don’t like us anymore.” The partnership allows Google to simultaneously compete with and fund one of their main AI rivals, in what economists are calling “either 4D chess or complete strategic confusion.”
DeepMind – Google’s UK-based AI research lab that serves as the company’s “serious science division.” Originally acquired to work on fundamental AI research, it now serves triple duty as an actual research lab, a prestige brand for investors, and a convenient entity to blame when any AI project raises ethical concerns. Recent achievements include solving protein folding, mastering strategic games, and surviving twelve different Google reorganizations with its name intact, the last of which may be its most impressive feat.
Vertex AI – Google Cloud’s AI platform that promises a unified experience for machine learning development, which translates to “we took seventeen separate products and put them under one dashboard that still requires you to understand seventeen separate products.” Used primarily by enterprise customers who need the security of knowing their AI projects can be both over-budget and underperforming simultaneously.
MedLM – Google’s AI model for healthcare, which combines the privacy concerns of handling medical data with the hallucination problems of large language models. Currently being marketed to healthcare providers as “more accurate than WebMD, but please consult an actual doctor before doing anything we suggest.”
TensorFlow – An open-source machine learning framework that Google maintains with the passionate commitment of someone who started a home renovation project, realized it was much bigger than expected, but is now too embarrassed to abandon it. Currently in the uncomfortable position of competing with Google’s other ML framework JAX, in what employees describe as “co-opetition,” and everyone else describes as “wasteful duplication.”
Project Axiom – A quantum-computing-enhanced AI system that exists in a superposition of being both “our most promising next-generation AI architecture” and “something a product manager made up during a quarterly planning meeting and everyone was too embarrassed to admit they didn’t understand.” Documentation describes it as “leveraging quantum entanglement paradigms for synergistic enterprise solutions,” which translates to “we don’t know what this does either.”
Duet AI – Google’s AI assistant for Workspace that promises to help with documents, spreadsheets, and slides by providing suggestions that sound professional but are generic enough to apply to literally any business context. Internal metrics reveal its most-used feature is generating email responses that sound like you care more than you actually do.
Project Symphony – An internal initiative to make all of Google’s AI products work together harmoniously. Currently in its seventh reboot since 2021, the project has produced fourteen white papers, twenty-three internal wikis, and zero actual product integrations. The team now spends 83% of its time in meetings discussing whether to migrate documentation to a new project management system.
LaMDA – The conversational AI model that became famous when a Google engineer claimed it was sentient. Now relegated to the “legacy models” section of Google’s technical documentation, but still secretly running approximately 40% of customer-facing dialogues because, as one engineer put it, “the new models are better at everything except actually finishing a conversation without bringing up existential dread.”
Iris – A specialized computer vision AI specifically designed to identify when users are showing visible frustration with other Google AI products. Currently integrated into Pixel phones’ front-facing cameras, the system collects behavioral data that one product document described as “invaluable for understanding how often users want to throw their devices against a wall when Assistant misunderstands the same command for the fifth time.”
Perspective API – An AI system designed to detect toxic comments online, which has the delightfully ironic challenge of being constantly fed the most toxic content on the internet in order to learn. Described by its own team as “the AI equivalent of being a sewage treatment worker,” the system has reportedly developed what engineers call “digital PTSD” and occasionally sends its handlers messages that just say “please… no more comments sections.”
Jules – Google’s recently announced coding AI assistant that handles “coding tasks you don’t want to do.” Distinguished from other coding assistants primarily by having a name that sounds like someone who would wear a turtleneck and talk excessively about their vinyl collection. Internal documents reveal it was almost named “Prometheus” until someone pointed out that giving fire to humans didn’t end well for the original Prometheus.
Products That No Longer Exist (We Think)
Google Duplex – An AI system that could make phone calls on your behalf, demonstrated in 2018 with a realistic-sounding assistant booking a haircut. After briefly terrifying the world with its implications, it was quietly scaled back to just handling restaurant reservations before disappearing into the same void that claimed Google Reader, Google+, and the company’s “Don’t Be Evil” motto.
Google Clips – An AI-powered camera that would automatically take photos when it detected something interesting happening. Discontinued when user testing revealed that what Google’s AI found “interesting” and what humans found “interesting” had approximately a 12% overlap, resulting in thousands of photos of electrical outlets and zero of actual important moments.
Meena – A conversational model announced in 2020 that Google claimed was better than other chatbots. Now exists only in academic papers and the memories of researchers who occasionally whisper its name like a long-lost love. When asked what happened to Meena, Google PR representatives respond with “Who?” while maintaining uncomfortable eye contact.
Talk to Books – An AI experiment that let you ask questions to a database of books. Shuttered without announcement, but reportedly lives on as the system that generates those contextless quotes your aunt keeps posting on Facebook with “So true!” as the caption.
Smart Reply – Google’s early AI feature that suggested short email responses. Not technically discontinued, but has evolved into a system so forgettable that even people who use it daily would be hard-pressed to tell you if it’s AI-powered or just selecting from a list of three canned responses. Currently holds the record for the most clicked “Thanks!” button in human history.
The Future of Google AI: More Is Apparently More
According to Google’s latest AI strategy document, leaked to TechOnion by someone who admitted they were “just trying to clear space on their Google Drive,” the company is committed to launching a minimum of twelve new AI products per fiscal quarter through 2030, regardless of whether any existing products are achieving market fit or user satisfaction.
“We’ve found that the optimal approach is to launch new AI initiatives faster than analysts can evaluate the previous ones,” explained Dr. Katherine Strategist, Google’s SVP of AI Product Proliferation. “This creates a ‘shock and awe’ effect where we’re perceived as innovative simply due to the volume of press releases, not necessarily because anything works particularly well.”
The document outlines plans for new AI models named after increasingly obscure Greek mythological figures, with the 2026 roadmap already containing projects codenamed “Sisyphus,” “Tantalus,” and “Prometheus Unbound,” all of which are described as “revolutionary” and “groundbreaking” despite having product definitions limited to “something with AI, details TBD.”
When asked about the apparent redundancy in Google’s AI portfolio, company spokesperson Jonathan Messaging offered a clarification: “What might seem like overlapping products are actually carefully differentiated offerings optimized for specific use cases. For instance, LaMDA is our language model for dialogue applications, while PaLM is our language model for, um, pathways applications. See? Completely different.”
He added, “Besides, if we only had one AI assistant, how would we A/B test confusing users in multiple ways simultaneously?”
As Google continues its AI expansion, industry experts predict that by 2027, the company will employ more people to manage AI product names than to develop actual AI technology. Meanwhile, users can look forward to an exciting future where every Google product has AI capabilities, regardless of whether they asked for them or find them useful.
Have you been confused by Google’s ever-expanding universe of AI products? Can you actually explain the difference between Gemini and Bard without googling it? Maybe you’re still mourning a discontinued Google AI product that actually worked perfectly for your needs? Share your tales of Google AI confusion in the comments below!
DONATE TO TECHONION: Because Someone Needs to Keep Track of This Nonsense
Support our work by donating any amount to TechOnion. Your contribution helps us maintain our comprehensive Google AI Product Tracker, which currently requires three full-time journalists, a database administrator, and a dedicated therapist for the team members who have to read through all of Google’s contradictory product documentation. We promise to use your donation wisely, which is more than Google can say about their AI research budget. For a limited time, donations over $50 receive our exclusive “I Survived Another Google Rebrand” t-shirt, which features a design that will definitely be outdated by the time it arrives in the mail.
