Cohere
Cohere: Frontier model, AI assistant, or foundation-model company shaping general AI interfaces and next-generation human-computer interaction.
ABAB Structured Brief
Cohere is indexed in ABAB Crypto Map under AI Models & Apps. This page keeps the official site, category, tags, and related ABAB coverage together as a searchable crypto project profile. Official domain: cohere.com.
Related News & Analysis
Cohere CEO Gomez: AI Deployment Requires Human Oversight and Ethical Considerations
Cohere CEO Aidan Gomez emphasized that in certain key areas, AI will always need "humans in the loop" for supervision and cannot be fully automated. He pointed out that both the public and private sectors should focus on...
Instaread: How Two Engineers Turned Reading into a 15-Minute Business
The central conclusion is that Instaread is not a company built around a single celebrity founder. It is the second-stage entrepreneurial project of two engineers who have worked together for years: Rahul Chitrapu and Vishnu Chapalamadugu. Instaread’s current product proposition is straightforward: bestselling books and other knowledge content are compressed into text or audio that can typically be consumed in about fifteen minutes. Its website now promotes book summaries, premium content from partners, and Instaread Originals, while the Apple App Store and Google Play listings emphasize thousands of titles, both audio and text, offline access, and original series. The founders are Rahul Chitrapu, who also appeared in earlier coverage as Rahul Simha or Rahul Simha Chitrapu, and Vishnu Chapalamadugu. Rahul’s publicly indexed LinkedIn profile still identifies him as Co-founder & CEO of Instaread. Vishnu is commonly listed by startup databases as Co-founder & COO. Third-party sources are not perfectly consistent about Vishnu’s exact title, so the safest characterization is that Rahul has generally served as the more publicly visible CEO-type co-founder, while Vishnu has remained the other core co-founder with a strong operating and organizational role. Judging from its publicly visible funding, team size, and media presence, Instaread is best understood as a durable but relatively lean vertical content-technology company, rather than a large publishing conglomerate or heavily financed technology unicorn. Wellfound currently places the company at 11–50 employees and approximately $1.7 million in total funding, although commercial databases disagree on the exact capital raised. Family background and early life are unusually absent from the founders’ public narratives; their public identities are almost entirely professional. For Rahul, reliable public information about date and place of birth, parents, family socioeconomic class, and childhood circumstances is limited / cannot currently be confirmed. The same is true for Vishnu. It would therefore be inappropriate to infer whether either came from a wealthy family, entrepreneurial family, academic household, or significant family capital. What can be established is that both had strong engineering and systems-oriented backgrounds. This matters because although both iDreamBooks and Instaread operate in publishing and content, their underlying approach has consistently resembled engineering: structure information, score it, compress it, automate parts of the workflow, and distribute the output efficiently. That continuity can be seen across both generations of products. Rahul Chitrapu had an engineering background at the University of Waterloo and nearly three years of documented professional experience at NOVA Chemicals before entrepreneurship. Rahul’s publicly indexed LinkedIn information lists the University of Waterloo under education, while Wellfound describes him as an engineer from the University of Waterloo. His precise field of study, degree title, and graduation year are publicly limited / cannot currently be confirmed. LinkedIn’s public search result also shows employment at NOVA Chemicals from September 2007 to July 2010, approximately two years and eleven months. The indexed information does not provide enough detail to establish his exact position, so it would be inappropriate to speculate further. Rahul therefore does not appear to have gone directly from university into a consumer-Internet startup. His verifiable trajectory is closer to engineering education → industrial corporate experience → book-information startup → mobile-content subscription startup. That background also helps explain why his businesses have tended to focus on data structures, automation, and efficiency rather than emerging from a traditional journalist, editor, or author career. The latter point is an analytical inference based on his career and products. Vishnu Chapalamadugu’s technical training is even easier to trace through academic records and clearly involved process control, energy systems, and engineering modeling. Vishnu’s public LinkedIn page lists Carnegie Mellon University as his educational institution, and Wellfound likewise identifies him with a Carnegie Mellon engineering background. Carnegie Mellon process-systems material also listed Vishnu as an M.S. student arriving from India’s Dr. B. V. Raju Institute of Technology and working with Professor B. Erik Ydstie. This confirms that he entered a graduate-level process-systems research environment at CMU. The exact final degree and completion date are not currently confirmable from the available public sources. More importantly, his technical background was substantive rather than merely a university affiliation. At the 2008 ASME Power Conference, Vishnu co-authored research involving sensitivity analysis and control of a gasifier with recycled CO₂; his affiliation in the paper included NOVA Chemicals in Sarnia, Ontario. In 2009, he co-authored work with Kendell R. Jillson and B. Erik Ydstie on inventory and flow control for an Integrated Gasification Combined Cycle process with CO₂ recycling. Vishnu’s early training therefore appears to have centered on systems modeling, dynamic control, and optimization of complex processes. His later move from engineering into Internet publishing appears dramatic on the surface, but at a more abstract level the continuity is strong: both involve decomposing complex systems into structures that can be measured, processed, and optimized. This is an inference from his record, not a stated personal philosophy. The real prehistory of Instaread is iDreamBooks. Without understanding iDreamBooks, it is difficult to understand why Instaread exists. In 2012, Rahul Simha, Vishnu Chapalamadugu, and Mohit Aggarwal founded iDreamBooks in San Francisco. Its simplest description was a “Rotten Tomatoes for books”: the service aggregated professional criticism, blogs, and user ratings and turned them into a more easily understandable evaluation system for deciding whether a book was worth reading. Publishers Weekly reported that the service aggregated more than 3,000 sources and had a staff of about five. The underlying problem definition is important. The founders were not trying to “publish more books.” They were addressing information overload and discovery efficiency in the book market: when readers face an enormous volume of books, how can they quickly determine what deserves their time? This was already the precursor to the Instaread problem. iDreamBooks answered “help readers choose”; Instaread would later answer “read, analyze, and compress it for them first.” iDreamBooks also attracted an unusually relevant early backer: Patrick Lee, co-founder of Rotten Tomatoes, became an early investor. The startup also received backing from 500 Startups. Business Standard reported that the initial money had come from family and friends before the company attracted 500 Startups funding, while Publishers Weekly independently confirmed the Patrick Lee and 500 Startups connections. The value of those relationships was not merely financial. 500 Startups, Patrick Lee, and the Silicon Valley startup network gave two engineers social capital, credibility, introductions, and access to the U.S. consumer-Internet and publishing markets. The most important commercial validation for iDreamBooks was its Sony Reader Store partnership, but that relationship also exposed the structural limitations of the first business model. In April 2013, Sony’s Reader e-bookstore integrated iDreamBooks ratings. Consumers viewing books in the Sony store could see iDreamBooks review and rating information. Publishers Weekly reported it as a formal partnership, while iDreamBooks’ own old blog explained that the integration operated through its API. The business model was already becoming clear. Business Standard reported that the Sony relationship generated an annual fee and that iDreamBooks regarded data licensing as an important source of revenue, targeting publishers, retailers, libraries, and other book-discovery businesses. In other words, the first business was primarily B2B data/API infrastructure for discovery. The weakness was that this model depended heavily on third-party book retailers and discovery platforms while requiring continuous expansion of review sources and database coverage. iDreamBooks was not necessarily a destination consumers had to use every day; it functioned more like a ratings layer embedded into other platforms. An external event later illustrated that dependency. Sony officially announced that the Reader Store in the United States and Canada would close on March 20, 2014, with customers transferred toward Kobo. Sony’s own FAQ confirms the closing date and migration plan. It would be too strong to claim that Sony’s closure directly “caused” Instaread; Pear’s later account does not say that. However, when the loss of that distribution channel is considered together with Pear’s statement that the founders explored several publishing ideas before deciding summaries were the best way to distribute knowledge broadly, 2013–2015 can reasonably be interpreted as the transition from helping readers discover books to producing compressed knowledge products directly for them. The shift from iDreamBooks to Instaread was not merely a rebranding exercise. It moved the founders to a different position in the publishing value chain. Pear VC’s 2016 investment announcement remains one of the most informative first-party accounts of the founders’ early Instaread period. Pear partner Mar Hershenson met Rahul and Vishnu by chance during a break at the PostSeed Conference. Pear initially made a relatively small investment so it could help the company and get to know the founders better. Pear specifically described the pair as exceptionally “scrappy,” recounting that they loaded their belongings into a car and drove from Toronto to San Francisco to pursue the startup. Early accounts also place them in an extremely small office on Castro Street in Mountain View. The significance is structural: they did not enter publishing backed by an established media conglomerate. They were classic engineer-founders using accelerators, seed investors, and Silicon Valley networks to gain entry. Pear also stated explicitly that the founders had tested several ideas in publishing before concluding that summaries were the best way to distribute knowledge to the largest number of people. That decision moved them beyond supplying book-rating data and toward controlling more of the process: selection, reading, analysis, writing, editing, audio production, and subscription distribution. The difference can be reduced to two sentences: iDreamBooks: “We tell you which book is worth reading.” Instaread: “We read it first and compress the most important material for you.” The first sells discovery and decision efficiency. The second sells time savings and knowledge-acquisition efficiency. Instaread’s funding was modest, but the quality of its network was meaningful, and public databases disagree about the exact amount raised. Wellfound currently records three early financing rounds: approximately $325,000 in September 2014, $250,000 in August 2015, and $1.1 million in May 2016, for roughly $1.675 million in total. Company profiles typically round this to $1.7 million. Jane Friedman’s 2017 publishing-industry discussion of Instaread likewise described three rounds totaling around $1.7 million. Pear VC publicly announced its investment in 2016 and described an initial smaller investment followed by a deeper, longer-term relationship. Pear’s account also said Ajay Kamat joined the company’s board. Funding databases nevertheless contain inconsistent figures. Wellfound and Jane Friedman converge around $1.7 million, while other commercial databases have historically shown figures around $1.66 million or as high as roughly $2.39 million. Therefore, approximately $1.7 million is best treated as the most commonly supported public figure rather than an audited absolute total. The current valuation, cap table, and founder ownership percentages are publicly limited / cannot currently be confirmed. The capital network should also be viewed across both ventures. The iDreamBooks period included 500 Startups and Patrick Lee; the Instaread period added Pear VC. Wellfound currently lists Patrick Lee, Richard Wolpert, and Andy Agrawal as advisors, although a platform listing alone does not establish how active those advisory relationships remain today. There is therefore no obvious story of Instaread being controlled by a major publishing conglomerate or financial group. Its more important background resource has been a network of Silicon Valley seed investors, accelerators, consumer-Internet founders, and technical talent. Instaread’s core business model monetizes a specific scarcity: the user’s lack of time. What customers are fundamentally purchasing is not an e-book but a compression ratio: material that might take hours or days to read is transformed into an approximately fifteen-minute text and audio knowledge unit. The website calls these “Key Insights from Bestsellers,” while the Apple and Google listings emphasize thousands of titles, fifteen-minute consumption, text and audio, offline access, and a personal library. Commercially, this is more direct than iDreamBooks’ API licensing because users pay Instaread itself. The current U.S. Apple App Store listing displays subscriptions of $8.99 per month or $89.99 per year, each with a one-week free trial. Instaread’s own Terms of Use also describes a seven-day free trial for new subscribers. Prices can vary by region, channel, promotion, and time, so these should be interpreted as prices displayed by the current U.S. Apple listing rather than universal pricing. The strategic change is significant: iDreamBooks primarily sold to potential retailers, publishers, and discovery platforms; Instaread primarily sells to end readers. The model therefore shifted from B2B information infrastructure toward a B2C recurring-subscription content business. Instaread’s most important assets are not the personal fame of its founders but the content-production and distribution system accumulated over more than a decade. The first layer is the content library. The current applications claim thousands of professionally written titles as well as Instaread Originals, including short-form material on people, business, politics, science, and other subjects. The second layer is its multimedia production pipeline. Apple and Google describe a process involving writers, editors, voice actors, artists, and fact-checking. Instaread’s Submittable page further states that contributors are compensated, summaries are edited for accuracy, and content is converted to audio. This confirms that Instaread is not merely an open user-generated-content platform; it operates an editorial, review, and audio-production workflow. The third layer is distribution infrastructure: the Instaread website, iOS app, Android app, and subscription-account system. The website additionally links to Teams, an Instaread Player, a WordPress Plugin, and a Newsletter, indicating that the business is not confined to a single mobile application. The fourth layer is a brand-extension media property, The Nugget, positioned as “a place of inspiration and learning, by Instaread.” It covers self-help, work, life, society, history, and news. New posts were still appearing in September 2026, indicating that it remains an active content-discovery and audience-acquisition property rather than merely an abandoned legacy blog. The fifth layer is partner content. The 2026 homepage explicitly promotes “premium content from our partners.” The full partner roster, revenue-sharing arrangements, and underlying licensing agreements are publicly limited / cannot currently be confirmed. It would therefore be incorrect either to assume that all partner content represents traditional licensing or to assume that all summaries are unlicensed. In asset terms, Instaread’s most meaningful value is likely in brand, content inventory, production processes, mobile distribution, subscriber relationships, and accumulated content metadata, all of which are operating intangible assets. Investor and advisor relationships with 500 Startups, Pear, Patrick Lee, and others are better understood as network and influence assets. This is a business-analysis framework, not an accounting classification from Instaread’s balance sheet. Its business-model evolution can be understood as three successive attempts to move closer to the economic value of the reader’s time. The first stage, iDreamBooks, performed judgment compression: thousands of reviews became an easier-to-read signal about book quality. The second stage, Instaread, performs content compression: entire books are transformed into fifteen-minute key-insight products sold through subscription. The third stage moves toward a broader short-form knowledge media business. Instaread Originals, “Success Stories,” “Short Cuts,” premium partner content, and the continually updated Nugget suggest an attempt to make Instaread something users visit continuously rather than only when they are considering a particular book. That shift matters commercially because a pure book-summary utility can be relatively low-frequency. Expanding into original material, topical subjects, influential people, and news interpretation potentially creates more reasons for users to return. This is a strategic inference from the current product architecture, not a disclosed financial result. The most consequential decisions and strongest achievements were not about fundraising; they were about repeatedly moving to a more defensible position in the value chain. The first important decision was to enter book discovery rather than conventional publishing in 2012. That allowed two engineers to begin with aggregation, scoring, algorithms, and APIs without first owning a publishing house, author roster, or large editorial organization. The second was to enter the U.S. startup ecosystem through 500 Startups and Silicon Valley networks. That gave iDreamBooks access to a sector-relevant investor such as Patrick Lee and to a corporate partner such as Sony. The third and most important decision was to move beyond the discovery layer and manufacture content directly. Pear’s account makes clear that after testing multiple publishing ideas, the founders decided summaries offered the strongest format for distributing knowledge. That move took them from a data feature that could potentially be replicated by Goodreads, Amazon, or a retailer into a product that could own the consumer relationship and recurring subscription revenue. The fourth was to make the product audio as well as text. That fundamentally expanded the use case: content could be consumed while commuting, exercising, or doing household tasks rather than requiring dedicated reading time. Apple and Google still treat this as a core product benefit. The fifth was to expand into Originals, short topical formats, and The Nugget, moving toward a higher-frequency knowledge brand rather than remaining solely a book-summary utility. Instaread’s strongest achievement is not that it “transformed the entire publishing industry”; the public evidence does not justify such a claim. A more defensible assessment is that it survived the early-2010s publishing-tech startup cycle and turned a relatively lightly funded venture into a subscription knowledge product spanning Web, iOS, and Android, with thousands of pieces of content and more than a decade of continuity. The U.S. Apple App Store currently shows roughly 9,300 ratings and a 4.6/5 score. Instaread’s website says it is “Used by millions,” while Pear stated in 2016 that millions of Instareads had already been read or listened to. Those “millions” figures should be treated as company/investor marketing or usage claims, not audited paid-subscriber or monthly-active-user figures. The main controversies concern copyright, fair use, and the quality limits inherent in commercial book summaries, rather than a major personal scandal involving the founders. Publishing-industry analyst Jane Friedman devoted a 2017 article to Instaread titled The Curious Case of Instaread: Copyright, Fair Use, and Rights Holders. Her basic description was that Instaread used writers to produce nonfiction-book summaries. The framing captures the central publishing-industry question around the model: to what extent may a commercial company summarize, analyze, and monetize copyrighted books without following the same rights-acquisition model as a conventional publisher? It is essential to distinguish a copyright debate from a legal finding of infringement. Reliable public sources establish that Instaread’s business model has been discussed through the lenses of copyright, fair use, and rights holders. But public information establishing a final court ruling against Instaread for the model, major damages, or a major publicly disclosed settlement is limited / cannot currently be confirmed. The publishing controversy should therefore not be presented as proof of unlawful conduct. The model also contains an inherent editorial risk. A fifteen-minute summary requires aggressive compression, and compression inevitably involves selection, interpretation, and framing. Instaread attempts to mitigate this through professional writers, editors, fact-checking, and multi-role production, but the question of whether a summary should complement or substitute for the original book remains structurally unresolved. There have also been product-experience criticisms. A historical Apple App Store review praised the content while criticizing aspects of the audio player and app experience; Rahul personally replied as a founder, asking about the user’s version and saying some issues had been addressed. This is evidence that UX execution has faced criticism at points in the company’s history, but it does not establish that identical problems persist today. The current overall iOS rating remains 4.6/5. As of 2026, Instaread’s real-world position can be summarized as follows: it did not become a giant publishing platform, but it has become a durable independent “knowledge-compression” brand. The main website remains operational in 2026 and continues to display current content. The Nugget published new pieces in September 2026 across politics, history, fiction, and self-improvement, and both Apple and Google storefront listings remain available. Google Play’s public page shows the Android app as last updated on August 30, 2023, so the Android update cadence does not appear particularly aggressive, while the web-content operation is visibly active in 2026. Rahul’s indexed LinkedIn profile still describes him as Co-founder & CEO at Instaread, and his X profile continues to identify him as Co-founder @instareads in San Francisco. Vishnu’s public professional profiles also remain tied to Instaread. Structurally, what they created is not primarily a personality-driven thought-leadership brand. It is a form of productized knowledge intermediation: long-form intellectual products created by authors are selected, interpreted, reorganized, edited, narrated, and transformed into secondary knowledge products optimized for mobile-era attention spans. That makes the founders’ long-term professional theme unusually coherent: First solve “Which book deserves my time?” Then solve “How can I obtain the core value of the book with less time?” iDreamBooks addressed the first question; Instaread addresses the second. Viewed as a timeline, the founders’ development is particularly clear. Around 2007–2010: Rahul worked at NOVA Chemicals. Vishnu was active in Carnegie Mellon/process-control research and industrial engineering environments, contributing in 2008–2009 to research on IGCC, gasifiers, and control systems. 2012: Rahul, Vishnu, and Mohit Aggarwal founded iDreamBooks, transplanting the Rotten Tomatoes aggregation-and-scoring concept into the book market. 2012–2013: The company received initial family-and-friends capital, support from 500 Startups, and resources from investors such as Patrick Lee while developing its review-aggregation and book-rating system. 2013: Sony Reader Store integrated iDreamBooks, providing a real B2B case for API and data licensing. 2014: Sony’s U.S. and Canadian Reader Stores closed. In the same year, startup financing databases show an approximately $325,000 early Instaread financing, indicating that the new venture was already taking substantive form. 2015: Instaread raised roughly another $250,000 and increasingly established its identity around mobile consumption and fifteen-minute book insights. Commercial databases differ on whether the official founding year should be recorded as 2014 or 2015. 2016: A roughly $1.1 million seed round was recorded. Pear VC publicly announced its investment and told the story of the founders driving from Toronto to San Francisco and testing various publishing concepts before choosing the summary model. 2017: Jane Friedman examined Instaread through the framework of copyright, fair use, and rights holders, making copyright boundaries one of the company’s most visible external controversies. Later years: The product expanded beyond a single book-summary format into Originals, profiles and topical content, audio, The Nugget, and premium partner content. 2026: The main website, subscription product, and The Nugget remain active. From this perspective, the most notable fact is not how much capital the company raised but that a venture funded at only a low-single-digit-million-dollar scale has maintained the product for more than a decade. The final assessment is that Rahul Chitrapu and Vishnu Chapalamadugu are best understood as engineer-founders in publishing technology—not traditional publishers and not personality-driven knowledge influencers. Their defining skill has not been writing a particular bestseller. It has been repeatedly identifying the commercial interface between information density and the user’s limited time. At iDreamBooks, they saw fragmented book criticism and compressed judgment into ratings. At Instaread, they saw long reading times and compressed books into fifteen-minute knowledge products. Adding audio converted otherwise unavailable commuting, exercise, and household time into potential consumption time. Adding Originals, The Nugget, and partner content represents an attempt to move from “summarizing a particular book” toward “continuously supplying compressed knowledge.” That strategic line is remarkably consistent from 2012 through 2026. Their strongest achievement has been reframing a publishing problem as an efficiency problem and constructing a repeatable production system around that insight. Their most obvious structural weaknesses are the copyright boundaries of summary products, information loss caused by extreme compression, product commoditization, and the unresolved question of whether customers use summaries to complement books or replace them. The first conclusion is supported by their product and investment history; the latter issues are structural risks of the model, not accusations of personal misconduct by either founder. On personal wealth and power, the evidence requires restraint. There is no reliable public basis for describing either Rahul or Vishnu as possessing enormous personal wealth, nor for calculating their net worth, ownership percentages, or Instaread’s current valuation. Public information is limited / cannot currently be confirmed. From the perspective of entrepreneurial durability, however, they have accomplished something less glamorous but more informative: since 2012 they have repeatedly built within the same “books—knowledge—time efficiency” problem space, converted the experience and vulnerabilities of the first venture into a second product, and kept that second product operating into 2026. That continuity is the key to understanding both Instaread and its founders.
The Transformer Revolution: How Eight Inventors Rewrote AI Architecture and Power
Scope of Invention The first thing to clarify is who “invented the Transformer.” By the strictest public-document standard, the Transformer was not the solo invention of one person. It was a collective invention by the eight coauthors of the 2017 paper Attention Is All You Need: Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan Gomez, Lukasz Kaiser, and Illia Polosukhin. The paper’s footnote explicitly says “Equal contribution. Listing order is random,” and then explains each person’s role in detail. That footnote matters a great deal, because it directly rules out the popular simplification that only the first author should count as the “real” inventor. The invention did not emerge from nowhere. It appeared at a moment when sequence transduction was hitting real bottlenecks. The dominant approaches at the time were RNNs, LSTMs, GRUs, and encoder-decoder systems augmented with attention, but those systems were either hard to parallelize, inefficient over long dependencies, or limited by long computational paths. What made the Transformer radical was not that it “discovered attention” for the first time; it was that it pushed the design to its logical extreme by removing recurrence and convolution entirely and letting self-attention become the central computational primitive. That made the model far better suited to massively parallel hardware and reshaped how later large models could be trained and scaled. The original paper’s results were not just interesting; they were decisive. It reported 28.4 BLEU on WMT 2014 English-to-German and 41.0 BLEU on English-to-French, while the larger model trained in 3.5 days on eight P100 GPUs and the base model trained in about 12 hours. So the Transformer was not only more accurate; it was also cheaper to train and easier to scale. The later large-model boom was built first on trainability and systems efficiency, and only then on the visible product layer. Even its naming and framing carried a “generality ambition” from the beginning. Based on the contemporaneous Google blog post and later media reconstructions, the model was never treated as merely a translation trick. It was framed almost immediately as a general architecture that could transfer across tasks and modalities. In the August 2017 official blog post, the team already highlighted parsing and projected future use in images and video. In other words, the Transformer was born not as a narrow translation model, but as a scalable computational framework for learning. By 2026, the paper’s citation counts are in the range where databases disagree but all still indicate extraordinary influence. Google Research and the NeurIPS listing show more than 240,000 citations, while Semantic Scholar reports roughly 172,905. The discrepancy reflects database and indexing differences, not disagreement about significance. By any serious measure, it is one of the defining AI papers of the century. Portraits of the Eight Co-Inventors Vaswani’s trajectory looks like a classic path from engineer, to foundational researcher, to platform entrepreneur. Public interviews show that he is the son of an architect and a doctor, that he grew up in Oman and later moved to Nagpur at age 15, and that he was influenced both by Indian scientists and by the Microsoft founding story. After studying computer science at BIT Mesra, he worked in IT, then left industry for graduate study at USC, where he completed a master’s and then a PhD in 2014 on statistical machine translation. The decisive intellectual shift in his story was not allegiance to one guru, but his recognition that deep learning was where the next real breakthroughs would happen. Professionally, Vaswani made his key leap at Google Brain, where he moved from statistical machine translation and NLP into much more general architectural design. The paper’s footnote states that after Jakob proposed the self-attention-over-RNN direction, Ashish and Illia designed and implemented the first Transformer models, and it emphasizes that Ashish was involved in nearly every aspect of the work. After the paper, he co-founded Adept in 2021 and then Essential AI in 2023, where he became CEO. Adept focused on models that take software actions, while Essential emphasizes enterprise AI systems and open-science frontier models. Shazeer’s public image is that of an unusually strong systems researcher with strong product instincts. Public information about his family background is limited, but his career path is clear: he graduated from Duke, joined Google in 2000, improved spelling correction for search, and later worked on core ad systems. By the time of the Transformer paper, he was already one of the most senior contributors in the group. His own website explicitly credits him with multi-head attention, the residual architecture, and the first superior implementation; Google Research now lists him as Gemini co-tech-lead. He co-founded Character.AI in 2021, then returned through the Google-Character licensing-and-rehire arrangement in 2024, and in 2026 he was elected to the U.S. National Academy of Engineering. Parmar’s story is almost the opposite of the standard elite academic pathway. She grew up in a lower-middle-class family in Pune; her mother had once wanted to become an architect but could not pursue that path, and that unrealized ambition pushed her to support her daughter’s own. Parmar did not get into IIT, turned instead to self-teaching AI, and when she first arrived in the United States for graduate study, her father and uncle had to borrow money to keep her afloat. Public reports differ on the exact name of her undergraduate institution: NDTV renders it as Pune Institute of Technology, while Forbes India says Pune Institute of Computer Technology. What is clear is that she completed a master’s in computer science at USC from 2013 to 2015 and then joined Google. Parmar’s role in the invention was far more substantial than the common “third author” shorthand implies. The paper’s footnote says she designed, implemented, tuned, and evaluated “countless model variants” in both the original codebase and Tensor2Tensor. That means she was not merely packaging results or helping with paper writing; she was central to turning an unstable invention into a scalable research program. She joined Google at age 24 as one of the youngest members of the team and one of the only contributors without a PhD, later co-founded Adept, served as its CTO, co-founded Essential, and by 2025 had moved into a technical role at Anthropic. Her long-term importance lies not only in symbolism, but in extending the Transformer into vision, audio, and 3D settings. Uszkoreit is the person who looks most like the group’s high-level architectural designer. Unlike many AI founders, he came from a household that was already deeply computational and linguistic: in his a16z interview, he says his father was a computer scientist and computational linguist and that dinner-table discussions included Turing machines and finite automata. What matters most is that Google Translate convinced him that machine learning could be both scientifically difficult and immediately product-relevant; that realization pulled him decisively back into Google. Publicly available material is much clearer on his career than on the full details of his degrees, but that career is unmistakable: Google Translate, Google Assistant semantic parsing, Google Brain Berlin, and then Inceptive. In the original invention, Uszkoreit’s most important role was directional. The paper explicitly says he proposed replacing RNNs with self-attention and started the effort to evaluate the idea. Later, he was also the author of the official Google blog post that introduced the model publicly. Afterward, he carried the same worldview into biology by founding Inceptive, which applies deep learning and experimentation to RNA and what he calls “biological software.” That continuity reveals his structural role: he is not merely an algorithm tinkerer, but someone who repeatedly searches for new domains where the “sequence-representation-generation” logic can dominate. Public information on Jones’s private background is relatively sparse, but his educational and career path is clear. He comes from a Welsh/U.K. background, completed a BSc in AI and Computer Science and an MSc in Advanced Computer Science at the University of Birmingham, and said in the university’s alumni material that the school’s reputation substantially helped him get into Google even without a referral. Professionally, he spent more than a decade at Google before co-founding Sakana AI with David Ha and Ren Ito and becoming its CTO. Jones’s contribution to the Transformer was also very concrete. The paper says he handled the initial codebase, efficient inference, visualizations, and ongoing model-variant experimentation. He was the kind of person who helps turn an elegant paper idea into a real research system: something that can run, compare, ablate, and convince others. That same character is visible in Sakana’s later direction, which is less about building a mass-market chatbot and more about running a research-first lab with a distinct Tokyo and partially open-source identity. Gomez was the youngest of the eight and one of the earliest to convert Transformer-era scientific influence into an enterprise platform. Public sources show that he was an undergraduate researcher at the University of Toronto, worked with Roger Grosse, interned and researched at Google Brain, and collaborated across both student and senior researcher settings. His personal website explicitly states that he was an undergraduate student of Roger Grosse, an intern of Łukasz Kaiser and Geoffrey Hinton, and later a doctoral student of Yarin Gal and Yee Whye Teh at Oxford. On the family side, a McKinsey profile says his parents deeply encouraged learning, and that his mother was British, studied dance, and became a librarian after moving to Canada. That combination of technical and humanistic input helps explain why his later company narrative consistently emphasizes the human side of AI. Gomez made two especially consequential decisions. The first was entering Google Brain at the undergraduate stage and moving directly from student researcher to co-inventor. CNBC still frames him in retrospect as a Google Brain intern who helped coauthor the paper that conceptualized the Transformer. The second was leaving the academic or quasi-academic path to co-found Cohere and anchor himself in enterprise AI rather than consumer chatbot hype. As for whether his Oxford doctorate was formally completed, public materials are not perfectly consistent: Oxford’s research group page long described him as a doctoral student, while LinkedIn shows a 2018–2024 study interval. Kaiser is the most clearly “theoretical computer scientist turned deep learning architect” among the eight. Public biographies say he was born in Wrocław, studied mathematics and computer science at the University of Wroclaw, completed his PhD at RWTH Aachen, and then worked as a tenured researcher in Paris on logic and automata theory before moving into Google’s semantic parsing work and later Google Brain. Public information on his family background is limited, but his intellectual formation is very clear: he entered modern AI not from product engineering but from logic, formal methods, and automata theory. In the Transformer project, Kaiser’s importance was infrastructural and organizational. The paper says he and Gomez spent “countless long days” building Tensor2Tensor, replacing the earlier codebase, improving results, and drastically accelerating research. Career-wise, he is distinctive because he did not quickly turn his fame into a startup brand. Instead, he remained in high-leverage institutional research. Public materials later place him at OpenAI, contributing to GPT-4 long-context work and appearing in 2025-era talks and papers as someone who co-authored Transformers and TensorFlow-level infrastructure. Polosukhin was the earliest among the eight to turn Transformer-era credibility into a decentralized AI infrastructure narrative. Public sources say he was born in Ukraine, studied applied mathematics and computer science at Kharkiv Polytechnic, moved to California after finishing his master’s, and then joined Google Research. Wired’s reconstruction is especially useful here: it describes him as working on direct-answer systems for Google Search, where the latency budget was brutally tight, which made efficiency and performance constraints central in his thinking. The paper footnote says that Polosukhin, together with Vaswani, designed and implemented the first Transformer models. But an equally important turning point came before the paper’s global fame fully arrived: he left Google in early 2017 and later co-founded NEAR Protocol in 2018. Today his public identity is no longer limited to “Transformer coauthor”; it is increasingly tied to decentralized, user-owned, verifiable, privacy-preserving AI. By 2026, business reporting depicts him as actively advocating AI agent infrastructure that is auditable and not excessively dependent on any one company. Collaboration Process and Turning Points The actual division of labor inside the paper is almost the full explanation for why the invention succeeded. Uszkoreit provided the central direction of replacing RNNs with self-attention; Vaswani and Polosukhin built the first working models; Shazeer introduced scaled dot-product attention, multi-head attention, and the key representational choices; Parmar and Jones expanded the search space through variants, tuning, code improvements, visualization, and inference; Kaiser and Gomez transformed the whole process through Tensor2Tensor. The Transformer, then, was not just “an idea.” It was the convergence of idea, implementation, systems engineering, tooling, tuning, and organizational coordination. That is also why the Transformer looks more like an industrial-research victory than a lone-genius breakthrough. Uszkoreit later described the project in precisely those terms: not as one overwhelming spark, but as the integration of prior attention work, optimizers, modeling judgment, implementation advances, and hardware-aware scaling. That observation is crucial because it explains why the most successful follow-on work came not from superficial paper imitation, but from labs that also had compute, systems, and research infrastructure. The publication timeline was also unusually compressed. The paper appeared on arXiv on June 12, 2017. Google’s official explanatory blog post followed on August 31, 2017. The paper then entered NeurIPS 2017. So the interval between “working internal result” and “publicly defining a new era” was only a matter of months. The Transformer was not a slow-burn idea; it accelerated through paper release, tooling, follow-on experiments, and adoption almost immediately. A compressed timeline looks roughly like this: 2017, the paper defines the architecture; from 2019 to 2021, the authors begin to split into differentiated organizational paths; in 2021 Adept is founded and Shazeer moves toward Character.AI; from 2019 through 2024 Cohere evolves from a high-profile research startup into an enterprise platform; in 2023 Essential and Sakana gain strong capital backing; and from 2023 to 2025 Inceptive, NEAR AI, Anthropic, OpenAI, and Gemini-related roles show how the original Transformer logic branched into biology, enterprise AI, open agent infrastructure, and frontier closed-model development. Organizations, Capital, and Business Models If you look only at the paper, these eight people are coauthors. If you extend the time horizon to 2026, they look more like an industrial network that radiated outward from Google Research and Google Brain into enterprise AI, consumer chat, frontier labs, bio-AI, Japan-based research labs, and decentralized AI infrastructure. Of the eight, Lukasz is the least startup-oriented in public form; the other seven all converted scientific prestige into some combination of companies, platforms, ecosystems, or investable organizational power. Vaswani and Parmar followed a path from research architecture to agentic software and then to enterprise foundation stacks. Adept aimed to make models take actions inside software rather than merely generate text. Reuters reported in 2023 that the company raised a fresh $350 million, bringing total funding to roughly $415 million. After leaving Adept, they co-founded Essential, which announced a $56.5 million Series A in 2023 with investors including Google, NVIDIA, AMD, and Thrive Capital. Their real asset is not only equity; it is the market’s belief that they can continue defining the next software substrate. Shazeer’s business path is closer to “research capability directly commercialized into conversational products and then partially reabsorbed by a tech giant.” Character.AI became one of the earliest major consumer products built around role-play and companion-style conversation at scale. Reuters reported that it had previously raised $193 million and reached a $1 billion valuation in 2023. The more consequential development was the Google licensing-and-rehire deal in 2024, which turned a single researcher-founder’s market value into something large enough to be discussed in multibillion-dollar strategic terms. Gomez’s business model matured earlier than many peers into a classic enterprise software path. Cohere did not define itself as “another ChatGPT”; instead it leaned into compliance, private deployment, long-term contracts, and workflow integration for businesses. Reuters reported in 2025 that annualized revenue had reached $100 million, that about 85% of the company’s business came from private deployments, and that valuation in different 2025 reports ranged from around $5.5 billion to $6.8 billion depending on timing and round. Its real asset is not just model weights, but trusted deployment architecture, enterprise channels, and governance posture. Uszkoreit’s Inceptive represents a different kind of commercial translation altogether: moving Transformer-era sequence intuition into RNA and therapeutic design. Public reporting says Inceptive first raised roughly $20 million in seed financing and then another $100 million in 2023 from backers including NVIDIA, Andreessen Horowitz, and Obvious Ventures. This is not an API business. It is a deep platform play built around experiments, biological sequence design, and generative modeling in life sciences. Jones’s Sakana emphasizes a research-lab identity, a Tokyo base, and selective open release. The company announced a $30 million seed round in 2024, framed its mission around nature-inspired intelligence, and quickly released Japanese models, some of them open. Its assets therefore include equity and team quality, but also a very distinct brand position: not a Silicon Valley clone, but a Japan-origin research-first alternative AI narrative. Polosukhin’s NEAR path is different again. NEAR Protocol is, on the surface, a blockchain network, but its current narrative clearly centers on NEAR AI, AI agents, privacy-preserving infrastructure, and user-owned AI. Its resource structure relies less on the classic VC-to-IPO path and more on protocol economics, ecosystem building, tokenized governance, and developer networks. For him, the real asset is not one product but an attempt to define a different ownership and trust model for the AI era. Kaiser’s situation is the most unusual. He did not bind his public identity to an independent startup. Instead, he embedded his value in research infrastructure and frontier-model work inside major organizations: TensorFlow, Tensor2Tensor, the Transformer, GPT-4 long-context contributions, and later reasoning-related work. People like this do not necessarily own famous product brands, but they often hold disproportionate influence over the internal direction of model systems and research programs. Achievements, Controversies, and Present Position The most impressive thing these eight people achieved was not just publishing a massively cited paper. They changed AI’s default building block. Before 2017, recurrence still looked like the natural default in NLP; after 2017, self-attention progressively became the dominant scaffold. And the architecture did not stop at language. It expanded into vision, music, code, biology, agents, and multimodal systems. Google’s own blog already hinted at image and video directions in 2017, and the authors’ later careers effectively became a human timeline of those expansions. The outside world remembers them today not because all eight names became universally famous, but because together they now occupy many of the key forks in modern AI: Shazeer on the Gemini and consumer-chat axis, Gomez on enterprise AI platforms, Vaswani and Parmar on agent automation and enterprise stacks, Uszkoreit on AI-biology, Jones on new research-lab models and Japanese AI work, Polosukhin on decentralized AI infrastructure, and Kaiser on frontier-model engineering. They are not merely historical figures; they are still actively shaping the field. The most visible public controversies around this group are concentrated in their later commercialization paths, not in the 2017 paper itself. Mainstream coverage has not centered on serious academic misconduct allegations regarding the paper. The recurring debates are instead about open versus closed development, consumer products versus enterprise deployment, and whether large incumbents are reabsorbing talent through licensing and deal structures. Google’s Character-related arrangement was reported in the context of broader scrutiny around how big tech acquires AI talent; Cohere has openly favored enterprise deployments over mass-consumer novelty; Vaswani has publicly argued for open science; and Polosukhin increasingly argues for user-owned, privacy-first, verifiable AI. The deepest argument is no longer over authorship. It is over who will control power in the Transformer era. Condensed to one sentence, the conclusion is this: the Transformer was not invented by one heroic genius, but by an eight-person team that simultaneously aligned theoretical judgment, implementation quality, tools, systems knowledge, organizational resources, and industrial ambition; and their later divergence now looks like a miniature map of the modern AI industry itself. If you want to understand today’s conversational AI, enterprise deployment, agent automation, RNA design, Japanese local models, decentralized AI, long-context systems, and reasoning models, many of those traces lead back to the same 2017 collaborative footnote.
Colossus Editor-in-Chief Jeremy Stern Says It's Hard to Imagine Mark Zuckerberg and Elon Musk Stepping Down
...most competitive person in what he does, ideologically more coherent and less chaotic than many figures in artificial intelligence. The voting rights structure was described in the show as majority control, and it is pra...
Dell Founder Michael Dell: Company Platform Launches Full Range of Mainstream Large Models Including Kimi K2.5, Llama, Grok
...d closed-source large models, including Kimi K2.5, Mistral, Cohere, Google Gemma, Meta Llama, Qwen, Nvidia Nemotron, Grok, Deepseek, and Phi. Users can access these models via dell.huggingface.co. This initiative covers ...