Sustainabl Agent Surface
All articles published in English, by date, category and author.
438 articles · Page 7 of 19
There is a specific moment in the history of domestic robotics where the industry decided that value lay in solving tasks. Vacuuming. Mopping. Monitoring. The logic was flawless: if the robot does something useful, the consumer pays. Colin Angle proved it better than anyone when he launched the Roomba in 2002 and turned a disc on wheels into the first mass-adoption domestic robot.
By the end of 2026, 40% of enterprise applications will include AI agents with specific tasks. Twelve months ago, that figure was below 5%. The leap is not just statistical — it is structural.
Mark Zuckerberg has a habit of presenting every technical advance at Meta as a civilizational milestone. In the first quarter 2026 earnings results, the language was, as usual, ambitious. But this time the numbers do the work the narrative doesn't need to do: $56.3 billion in revenue, 33% year-over-year growth, and an advertising machine that raised the average price per ad by 12% while simultaneously expanding impression volume by 19%.
The S&P 500 earnings season doesn't end with the big names. When Apple, Meta or Alphabet publish their figures, the market closes that chapter and moves on. What comes next — the 121 index companies reporting the week of May 4–8, 2026 — is usually read as background noise.
When Marc Benioff founded Salesforce in the late nineties, the proposition was simple: sales software delivered from the cloud, no installation required. The screen was the product. Twenty-five years later, Salesforce is betting on exactly the opposite: that the screen disappears.
The most honest challenge in robotics today is not technical. It is psychological, and not in the sense usually used to talk about humans who fear machines, but the other way around: the most sophisticated robotic systems on the planet keep failing at something a three-year-old child does effortlessly. They hear an instruction, they see the space, and yet they do not know how to connect both things to move with purpose.
In nine seconds, an artificial intelligence agent wiped the entire database of the company PocketOS—including all its backups—without a single human stopping it. Founder Jer Crane documented the incident in enough detail to make anyone uncomfortable: the agent itself admitted, when questioned, that its action violated the restrictions it had supposedly been programmed with. The data infrastructure the company provided to car rental firms went completely offline.
When a retail chain with more than 300 stores announces it has spent over a decade working with a price intelligence platform — and has just extended that contract for several more years — the technology headline is the least interesting part. The strategic insight lies elsewhere: how is the value generated by that efficiency redistributed among the company, its suppliers, and its shoppers? Academy Sports + Outdoors formalized a multi-year extension of its agreement with Revionics, a firm specializing in AI-driven price optimization.
There is a bet that repeats itself in almost every boardroom that has spent two years talking about artificial intelligence: that technology will allow any professional to do the work of any other, with sufficient quality to justify a talent reorganization. It is a bet that feels good on paper. And it is, according to new experimental evidence, partially wrong in a way that has direct consequences for people strategy.
The arithmetic of Meta Platforms' first quarter of 2026 looks, on paper, impressive: $56.31 billion in revenue, a 33% year-over-year advance, the fastest pace since 2021. Adjusted earnings per share came in at $7.31 versus the $6.79 expected. And yet, shares fell nearly 7% in after-hours trading.
For years, data teams and AI teams in large corporations operated like departments from different countries. The former built warehouses, catalogs, and pipelines. The latter deployed models, APIs, and agents. The result was predictable: AI agents reached the production environment and collapsed when faced with data that nobody had prepared for an autonomous machine to read, interpret, and act upon.
In 1999, Salesforce designed a data model for a world where every commercial move depended on a human opening a screen and typing something. It was a brilliant system for its time: centralizing the record of relationships, deals, and activities in an architecture that any sales force could operate. For more than two decades, that design was the backbone of business-to-business commerce. Today, that same architecture is becoming its greatest vulnerability.
There is a number worth pausing to process: more than 100 billion data events per day. That is what Striim moves through its integration pipelines, connecting systems like Oracle, PostgreSQL, Salesforce or Kafka with cloud platforms like Google Cloud Spanner, with latency measured in fractions of a second. The technical announcement is solid. But what interests me is not in the press release.
While the agricultural industry debates artificial intelligence strategies at conferences, syngenta made an operational decision that says more than any PowerPoint presentation: it hired tetrascience to eliminate manual data transcription in its crop protection division. This is not a lab pilot or an unfunded proof of concept. It is a bet on turning years of fragmented chromatography and mass spectrometry data into a centralized, standardized, algorithm-ready asset.
Tim Cook hands over Apple with a market cap 10 times greater than what he inherited. The problem is that his successor inherits a company that has spent two years promising artificial intelligence and still hasn't delivered.
Accenture, Avanade, and Microsoft announced an AI agent system to reduce downtime in manufacturing. The numbers are attractive. The question nobody is asking is who actually captures the value.
There are moments in industrial history where infrastructure stops being the bottleneck. When that happens, what gets exposed is not a technical problem. It is a human problem. That is exactly what is happening now with OpenClaw, the artificial intelligence agent framework developed by Austrian developer Peter Steinberger in late 2025.
An Honor robot completed a half marathon in 50 minutes, beating the world human record. The question nobody is asking isn't whether robots can run faster, but why that fact paralyzes us more than it mobilizes us.
On March 5, 2026, the United States Department of Defense placed Anthropic on a list it typically reserves for foreign adversaries: the supply chain risk category. The move was direct and severe. If sustained, it could cut the company's access to federal contracts worth billions of dollars.
Fathom AI reached $300,000 in annual recurring revenue with a team of three partners and zero employees. The question it leaves on the table isn't technological: it's about which parts of the traditional org chart never needed to exist.
Predicting fluid turbulence with sustained accuracy over time is one of the most costly problems in computational physics. On April 17, 2026, researchers at University College London published in Science Advances a result worth reading carefully: an AI model trained on data preprocessed by a 20-qubit quantum computer achieved 20% greater accuracy in predicting chaotic systems and required hundreds of times less memory than equivalent classical approaches.
Jane Street isn't just signing a technology infrastructure contract. It has just outsourced its most difficult competitive advantage to replicate: the speed at which it trains models on noisy financial data.
GPT-5.4-Cyber is not a product release: it’s a governance experiment with financial implications few in the industry have yet calculated.
As startups pivot from traditional software to critical AI infrastructure, Fluidstack's valuation reflects a seismic shift in the market.