Greg Isenberg outlines SF takeaways

Why it matters

Greg Isenberg said on May 27 that five days of meetings in San Francisco left him with a specific view of the AI market: companies are hiring for people who can get agentic systems working inside real customer workflows. In a post on X, the Late Checkout chief executive said he spent the trip with “frontier AI model teams, AI startup founders, and 3 billionaires,” and wrote that “forward-deployed engineer” had emerged as the hottest role in the conversations he heard. Isenberg’s thread tied that hiring demand to a broader shift in software. He said buyers were acquiring SaaS businesses and rebuilding them “agent-first,” while model companies were seeking more data on how products are actually used in the field rather than just API volume and token counts. ### Why did Isenberg focus on “forward-deployed engineers”? Greg Isenberg’s post described a market that wants engineers closer to deployment than research. (youtube.com) His notes said model companies can see “API calls and token counts” but “can’t see the actual workflows,” making operators with niche, on-the-ground usage knowledge more valuable. Inception, one of the companies cited in the editor’s source briefing as hiring in this category, describes its Forward Deployed AI Engineer role as sitting at the intersection of product engineering, customer implementation, evaluations, data collection, model optimization and enterprise deployment. (youtube.com) ### What does “agent-first SaaS” mean in his telling? Isenberg wrote that three billionaires he met were “buying SaaS companies and rebuilding them agent-first.” He said they were inspired by Bending Spoons and Ryan Cohen’s eBay deal, and summarized the playbook as buying the company, cutting headcount, rebuilding the technology, adding agents and features, and then raising prices. That description matches a product strategy centered less on a standalone chatbot and more on reworking software around automated task execution, workflow coverage and higher-value features. (roles.directory) In his post, Isenberg did not name the buyers he met or the specific companies they were targeting. ### Why did MCP endpoints come up so often? Isenberg wrote that “MCP came up in literally every conversation” and argued that companies exposing their products as MCP endpoints were being “pulled into deals they never pitched for.” He added: “The ones that aren’t are becoming invisible to agents.” (youtube.com) Greg Isenberg has separately described MCP, or Model Context Protocol, as infrastructure that lets services present tools in a standard way for AI systems to use. (youtube.com) In his earlier public material, he framed that as shifting integration work toward service providers and making products easier for agents to discover and call. ### Where do voice agents and open-source models fit? The social briefing tied Isenberg’s San Francisco notes to two other themes: voice agents and open-source models. It said his write-up flagged voice agents as a route to “the next billion users” and argued that open-source models such as Gemma and DeepSeek now cover about 80% of needs. Inception’s Mercury 2 launch points to why voice and agent loops are getting attention from builders. (luminarypodcasts.com) The company said on February 24 that Mercury 2 was built for “agent loops, real time voice and search,” with claimed throughput of 1,009 tokens per second on Nvidia Blackwell GPUs. ### What is “agent debt”? The social briefing said Isenberg also raised “agent debt,” a phrase used to describe the maintenance burden that appears after companies rush agentic features into production. (youtube.com) The post surfaced that concept alongside his hiring comments, suggesting the work is shifting from demos to iteration, monitoring and repair. May 27 is the key date for the thread itself, and the next public marker is whether the hiring language Isenberg highlighted appears more often in company job listings and founder posts. (inceptionlabs.ai) Inception’s forward-deployed role and Isenberg’s X thread are both live as of Wednesday. (roles.directory) (youtube.com)

Key numbers

• Greg Isenberg said on May 27 he returned from five days in San Francisco with notes that framed “forward-deployed engineers” and agent-first software as current hiring themes.
• In a post on X, the Late Checkout chief executive said he spent the trip with “frontier AI model teams, AI startup founders, and 3 billionaires,” and wrote that “forward-deployed engineer” had emerged as the hottest role in the conversations he heard.
• It said his write-up flagged voice agents as a route to “the next billion users” and argued that open-source models such as Gemma and DeepSeek now cover about 80% of needs.
• Inception’s Mercury 2 launch points to why voice and agent loops are getting attention from builders.

What happens next

• Greg Isenberg said on May 27 that five days of meetings in San Francisco left him with a specific view of the AI market: companies are hiring for people who can get agentic systems working inside real customer workflows.
• It said his write-up flagged voice agents as a route to “the next billion users” and argued that open-source models such as Gemma and DeepSeek now cover about 80% of needs.
• Inception’s Mercury 2 launch points to why voice and agent loops are getting attention from builders.

Sources

• youtube.com
• roles.directory
• luminarypodcasts.com
• inceptionlabs.ai

Quick answers