Sector 2 · 3 sub-themes
AI Agents & Tool Use
Agentic coding has crossed into the mainstream: Anthropic reports the share of GitHub projects with coding-agent activity has more than doubled since late 2025, with Claude Code driving most of that growth. Claude Code creator Boris Cherny says he no longer writes his own code. The frontier problem is now duration — an Anthropic engineer ran a 70-minute workshop on keeping agents on-task for hours or days without looping, drifting, or losing context. Underpinning this is the fast standardization around Model Context Protocol. Microsoft (Dynamics 365 Commerce), X, Navan, and Propel all launched MCP servers, exposing enterprise data to agents through one interface. That connective tissue is what makes long-running agents useful across systems. The craft is shifting accordingly. Practitioners are floating "loop engineering" as prompt engineering's successor: stop hand-crafting one-off prompts, start designing persistent, self-supervising agent workflows. SaaStr notes even mediocre prompts now yield decent software, lowering the value of precise manual prompting. The through-line: attention is moving from single turns to durable, tool-connected loops.
AnthropicClaude CodeBoris ChernyModel Context ProtocolPropelNavanMicrosoftDynamics 365 CommerceBusiness InsiderTech TimesSaaStragentic coding workflowsloop engineeringModel Context Protocol adoptionenterprise AI integrationautonomous agent orchestrationprompt engineering successoragentic scaffoldingenterprise authorization securityself-supervising agent loopscoding agent productivity
2.1
Anthropic Claude Code agents drive agentic coding adoption and workflow guidance
- Anthropic reported that the share of GitHub projects with coding agent activity has more than doubled since late 2025, with Claude Code driving much of this growth in agentic coding. [1]
- An Anthropic engineer hosted a 70-minute workshop specifically addressing how to keep agents running on-task for extended durations — from multiple hours to multiple days — without losing context, self-looping, or drifting from goals. [4]
- Boris Cherny, the creator of Claude Code, stated he no longer writes his own code, reflecting how agentic coding tools have changed the workflows of even their own developers. [3]
- VentureBeat reported that Anthropic announced a new feature for Claude Code users, described as potentially significant for how agentic coding workflows function. [2]
- The article behind source [5] (from "Making Developers Awesome at Machine Learning") draws a distinction — originating from Anthropic's "Building Effective Agents" piece — between agentic workflows (LLMs and tools orchestrated through predefined code paths) and autonomous agents (LLMs that dynamically direct their own process and tool use at runtime), warning that conflating the two leads to either over-engineering simple tasks or under-engineering open-ended ones. Deloitte projects that by 2027, up to 50% of companies using generative AI will have launched agentic AI pilots or proofs of concept. [5]
AnthropicClaude CodeBoris ChernyVentureBeatDeloitteGitHubMaking Developers Awesome at Machine Learningagentic codingautonomous agentslong-running agentsagentic workflowsdeveloper productivitygenerative AI adoptionLLM orchestrationcontext managemententerprise AI pilots
2.2
Model Context Protocol Gains Momentum as Enterprise AI Integration Standard
- Propel Software claimed to be the first company to launch a production MCP for Product Lifecycle Management (PLM), enabling AI agents to access and act on PLM data through standardized tool interfaces. [7]
- Navan launched an MCP server specifically for travel and expense management, allowing AI tools to interact with its platform's data and workflows through the Model Context Protocol. [3]
- Microsoft launched a Dynamics 365 Commerce MCP Server designed for AI agents, enabling agentic commerce use cases by exposing Commerce platform context and capabilities to AI tools via MCP. [6]
- X (formerly Twitter) launched an MCP server to make its platform more accessible to AI tools, allowing developers and AI agents to interact with X's data and features through a standardized protocol interface. [4]
- Marcora launched a free MCP tier that places company-specific context — internal knowledge and documentation — inside any AI tool, making enterprise knowledge accessible to AI agents via MCP at no cost. [2]
- EE Times reported that MCP is emerging as a common framework for enterprise AI systems, with broad adoption across industries signaling its consolidation as a de facto integration standard for connecting AI agents to business platforms. [5]
- Security Boulevard reported on the need to secure MCP deployments with quantum-resistant cryptography, reflecting growing enterprise concern about hardening the protocol as it becomes a critical integration layer for AI agents. [1]
Propel SoftwareNavanMicrosoftDynamics 365 CommerceMarcoraEE TimesSecurity BoulevardModel Context Protocol adoptionenterprise AI agent integrationMCP server launchesPLM AI integrationagentic commercequantum-resistant cryptographyAI tool interoperabilityenterprise knowledge managementMCP securityde facto AI integration standard
2.3
"Loop Engineering" Emerges as Proposed Successor to Prompt Engineering
- A concept called "loop engineering" is being framed as the successor to traditional hand-crafted prompt engineering, centered on designing autonomous agent workflows in which AI agents self-supervise and rework their own outputs rather than relying on human-crafted prompts. The approach is described as particularly prominent in AI coding contexts, with Claude Code cited as a specific implementation example. [1][3][4]
- Business Insider and Tech Times both characterize the shift as moving from one-off prompt crafting toward the design of persistent, iterative agent loops — the core idea being that practitioners "stop prompting, start designing" autonomous workflows. [3][4]
- Parallel coverage from Built In and Forbes suggests that prompt engineering itself is simultaneously being re-evaluated, with some sources positioning it as a skill that practitioners may "outgrow" as agentic tool use matures, while others continue to publish comprehensive prompt engineering technique guides. [2][5]
- SaaStr notes that even mediocre prompts can produce reasonably good software today, implying that the quality bar for manual prompt crafting has lowered as underlying models and agentic scaffolding improve — a datapoint consistent with the broader narrative that precise prompt engineering is becoming less differentiating. [6]
Claude CodeAnthropicBusiness InsiderTech TimesBuilt InForbesSaaStrloop engineeringagentic AI workflowsprompt engineering evolutionautonomous agent designAI coding assistantsself-supervising agentsiterative agent loopsagentic scaffoldingpractitioner skill shifts