IBM

Overview

IBM is a long-standing enterprise technology company with a significant presence in AI, open source, cloud, and enterprise infrastructure. In the newsletter, IBM appears primarily in connection with its Granite family of foundation models—especially Granite 4.1 3B—and with its role in agent interoperability through the donation of its ACP protocol, which later merged into A2A. IBM also surfaced through commentary from David Cox on open source, reinforcing the company’s ongoing engagement in practical AI development rather than only research branding.

For AI Product Managers, IBM matters because it sits at the intersection of model development, enterprise deployment, and standards. Its Granite models represent one path toward smaller, usable open models for real-world tasks, while its work around agent communication highlights how large vendors are influencing the emerging infrastructure layer for multi-agent systems. IBM is especially relevant when evaluating enterprise-safe AI stacks, open ecosystem participation, and the tradeoffs between model size, performance, and deployability.

Key Developments

• 2026-01-03 — DeepLearningAI featured IBM’s David Cox in discussion around open source wins, placing IBM in the broader conversation about how open approaches are shaping AI progress and adoption.
• 2026-02-12 — IBM’s agent communication protocol ACP was noted as having merged into A2A shortly after its donation. This positioned IBM as a contributor to emerging interoperability standards for AI agents.
• 2026-05-05 — IBM’s Granite 4.1 3B model was highlighted through Simon Willison’s prompting experiments generating SVG pelican images across different quantized variants. His takeaway was that output quality was mostly poor and did not show a clear relationship to model size, illustrating practical evaluation challenges for smaller open models.

Relevance to AI PMs

• Benchmarking open models for product fit — IBM’s Granite line is useful for evaluating whether smaller, more efficient open models are good enough for specific product workflows. AI PMs can use examples like the Granite 4.1 3B prompting tests to avoid assuming parameter count or quantization choice will reliably predict user-perceived quality.
• Planning for agent interoperability — IBM’s ACP contribution and its merge into A2A signal that protocol-level interoperability is becoming strategically important. AI PMs building multi-agent products, integrations, or orchestration layers should track these standards early to reduce future rework.
• Enterprise AI stack decisions — IBM is relevant in enterprise contexts where governance, infrastructure compatibility, and openness all matter. PMs comparing hosted APIs, open-weight models, and hybrid deployment options should include IBM in vendor and ecosystem scans, especially for regulated or large-scale business environments.

Related

• Granite 4.1 3B — IBM’s model family entry that appeared in Simon Willison’s prompting experiments.
• Simon Willison — Evaluated Granite 4.1 3B with a creative SVG generation prompt, offering a practical lens on model behavior.
• A2A / a2a-agent2agent-protocol — IBM’s ACP merged into A2A, linking IBM to agent interoperability efforts.
• David Cox — IBM leader referenced in discussion of open source wins.
• DeepLearningAI — Mentioned IBM via its coverage of David Cox and broader AI industry themes.
• Google Cloud — Related through the broader ecosystem of cloud and AI infrastructure players relevant to enterprise AI deployment and standards adoption.
• Linux Foundation — Connected through the open governance and standards-oriented ecosystem that often overlaps with interoperability and enterprise AI initiatives.