 |
|
|
|||||
As a consultancy owner, I’ve been experimenting heavily with the headline AI applications for the better part of two years now. Our teams have tested it across dozens of products and use cases. Some experiments worked immediately. Others failed at first but succeeded six months later when the models improved. Some we’re still figuring out. The results keep evolving. A lot of leaders are obsessing over AI strategies right now. Detailed roadmaps, implementation plans, and resource allocation. I get it. Leadership wants clarity, stakeholders want commitments, and everyone wants to know the plan. But here’s the issue. Technology is moving way faster than traditional planning cycles can handle. What seemed impossible in January becomes a commodity by June. GPT-4 launched in March 2023. By year-end, teams were already building multimodal AI and voice interfaces that didn’t exist when they started planning. So, we’ve developed a posture instead of just a strategy. WHAT DOES POSTURE MEAN? A posture is a consistent way of thinking about when, why, and how to experiment as things evolve. It’s the framework you use to make decisions in real-time when conditions keep changing. For us, that starts with a simple filter. Before we experiment with AI on any problem, we ask: Does this fit our criteria? We built a framework called SPARK to help us decide: Scale: High volume or time-intensive tasks Pattern: Repeatable structures or behaviors Ambiguity: Needs perspective or ideation Redundancy: Been done before, will be done again Knots: Bottlenecks that slow people down If a potential concept hits at least two of these markers, we move forward with an experiment. If not, we wait. Screening helps us focus on high-value opportunities instead of throwing spaghetti at the wall to see what sticks. WHY THIS COMPOUNDS OVER TIME Here’s what happens when you develop a clear posture: You get faster at recognizing valuable opportunities. You build institutional knowledge about what works in your specific context. You learn when to push forward and when to wait for technology to mature. One team we work with started experimenting with AI for customer support triage in early 2023. The initial results were mixed. AI frequently misrouted tickets and gave generic responses. Six months later, we came back to it. Better models, better prompting techniques, and a better understanding of what the AI could handle. This time it worked. They now process 60% of tier-one support interactions with AI, freeing their human team to focus on complex customer issues. The difference wasn’t a better strategy. It was having a posture that included “when to come back to something we already tested.” DEFINE YOUR OWN POSTURES You don’t need to copy our framework. Build something that fits your business context, risk tolerance, and team’s capabilities. But it may be helpful to think through these questions: What types of problems are we willing to experiment with? What results would make an experiment worth scaling? How do we balance speed with responsibility? What triggers a decision to invest more deeply or move on? How do we capture and share learnings across experiments? Having clear answers matters more than having perfect answers. THE LONG VIEW AI capabilities will only continue to evolve, and new use cases will emerge. Some of today’s cutting-edge applications will become commodities. Others will reach dead ends. I believe that the companies who will thrive will be the ones who can consistently evaluate new opportunities, learn from results, and adjust as conditions change. They’ll have trusted experts who know where to experiment and when to scale. That’s what I mean when I say our AI point of view isn’t a snapshot. It’s a posture. TL;DR The technology keeps moving. Our posture helps us move with it. George Brooks is the CEO and founder of Crema.
Category:
E-Commerce
Big Tech is on a spending spree, forecast to drop a staggering $650 billion on artificial intelligence (AI) in 2026 aloneand that’s just for Alphabet, Meta, Microsoft, and Amazon. The companies are ramping up their investment in an increasingly competitive, high-stakes arms race, pouring hundreds of billions into massive data centers and semiconductors, in hopes of establishing a long-term strategic advantage in their quest to dominate the future of technology. With all four reporting earnings within the last week, Wall Street’s reaction may be an indication that investors are increasingly worried about the large spend, and relative payoffs, from the AI investments. The spending also coincides with mass layoffs across the tech industry. Those layoffs, which were originally attributed to AI being able to do the jobs of human workers, are now being seen by critics as an excuse for companies to reduce headcounts, so companies can divert spending from workers to building and powering AI data centers, among other things. Here is a look at some at the numbers as we break down Amazon, Meta, Microsoft, and Alphabet’s AI spend for 2026. Amazon 2026 AI spend Reporting fourth-quarter earnings on Thursday, Amazon said it was pouring $200 billion in capital expenditures into AI this year. News of that, plus the fact it missed first-quarter operating income due to the massive spend, sent shares of the stock down 10% in early morning trading on Friday. At the time of this writing, shares of the cloud giant (AMZN) were down over 6% in afternoon trading. “With such strong demand for our existing offerings and seminal opportunities like AI, chips, robotics, and low earth orbit satellites . . . we anticipate strong long-term return on invested capital,” Amazon CEO Andy Jassy said in the earnings release. Alphabet 2026 AI spend Alphabet, Google’s parent company, said in Wednesday’s earnings report that it estimated AI spending would hit $175 billion to $185 billion this year. Despite its recent performance and positive earnings report, Wall Street reacted with caution, sending shares of Alphabet Inc. (GOOGL, GOOG) down nearly 2% at the time of this writing on Friday afternoon. Meta 2026 AI spend By comparison, Meta‘s capital expenditure for AI lags behind, but is significantly higher and more aggressive than just one year ago. The companywhich owns and operates Facebook and Instagram, as well as Threads, Messenger, and WhatsAppsaid it was hiking capital investment for AI development by 73% in 2026, to between $115 billion and $135 billion. For some context, at the beginning of 2025, Meta CEO Mark Zuckerberg had said the social technology company planned to invest between $60 billion and $65 billion, showing just how quickly this AI arms race has ramped up. Shares of Meta (META) were trading down less than 2% at the time of this writing on Friday afternoon. Microsoft 2026 AI spend Finally, Microsoft (MSFT)whose shares were up 1% Friday afternoon, bucking the trend of the three other Big Tech stocksis on course for AI capital expenditures of $145 billion by the end of its fiscal year in July, according to Yahoo Finance. The stock is down 41% from its October high.The company recently reported second-quarter 2026 earnings, including $81.3 billion in revenue (up 17% year-over-year), and diluted earnings per share (EPS) of $4.14 (up 24% year-over-year). We are only at the beginning phases of AI diffusion and already Microsoft has built an AI business that is larger than some of our biggest franchises,” Microsoft CEO Satya Nadella said in a statement. We are pushing the frontier across our entire AI stack to drive new value for our customers and partners.
Category:
E-Commerce
On Thursday, OpenAI released GPT-5.3-Codex, a new model that extends its Codex coding agent beyond writing and reviewing code to performing a much wider range of work tasks. The release comes as competition continues to heat up among AI companies vying for market share in the AI-powered coding tools space. OpenAI says GPT-5.3 combines the coding performance of GPT-5.2-Codex with the reasoning and professional-knowledge capabilities of GPT-5.2, while running 25% faster. This allows GPT-5.3-Codex to handle long-running tasks that involve research, tool use such as web search or database calls, and complex execution and planning across both general work tasks and software development. Codex has reached over 1 million developers, OpenAI claims. And while Anthropics Claude Code has also seen rapid adoption, head-to-head data comparing the two tools remains scarce. SemiAnalysis reports that 4% of GitHub public commits, or new code uploaded to repositories, are currently being authored by Claude Code, and projects that figure could reach 20% or more by the end of 2026. Benchmark one-upsmanship OpenAI says GPT-5.3-Codex now has the best score of any model on SWE-Bench Pro, which evaluates real-world software engineering across four programming languages. Same goes for Terminal-Bench 2.0, which measures the terminal skills coding agents need. Anthropic says its new Claude Opus 4.6 model, also announced Thursday, achieved top scores on several industry benchmarks including Humanity’s Last Exam (complex multidisciplinary reasoning), GDPval-AA (economically valuable knowledge work), and BrowseComp (hard-to-find information search). OpenAI says its new model is capable of taking into account larger bodies of information while working on a task, as well as thinking about those tasks for longer periods without human intervention. In testing, OpenAI says it saw GPT-5.3-Codex autonomously iterate on game development over millions of tokens using generic prompts like “fix the bug” or “improve the game.” Similarly, Anthropic says its new Opus 4.6 model can comprehend larger code bases and make more thoughtful decisions about how to add new code. OpenAI says GPT-5.3-Codex is built to support the full software lifecycle, including debugging, deploying, and monitoring code, as well as writing product requirement documents and conducting research. Beyond coding to knowledge work The same agentic capabilities that expand Codexs coding skill can apply to tasks well outside the realm of software development, OpenAI says, extending to things like creating slide decks, analyzing data in spreadsheets. On GDPval, an OpenAI evaluation measuring performance on well-specified knowledge-work tasks across 44 occupations, GPT-5.3-Codex matches GPT-5.2 while adding stronger coding capabilities. On OSWorld-Verified, which tests computer use in a visual desktop environment, GPT-5.3-Codex achieved 64.7% accuracy compared to 38.2% for its predecessor. Anthropic has taken its Claude Code tool in the same directionto help a wider pool of information workers with a far broader set of business tasks. GPT-5.3-Codex is the first model OpenAI classifies as “High capability” for cybersecurity-related tasks under its Preparedness Framework, and the first the company has directly trained to identify software vulnerabilities. OpenAI is committing $10 million in API credits to accelerate cyber defense, particularly for open source software and critical infrastructure systems. GPT-5.3-Codex is now available to paid ChatGPT subscribers in the Codex app, command line interface, IDE extension, and web. OpenAI says it is working to enable API access (used by enterprise and independent developers) to the model soon.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||