AI Agents Are About to Out-Consume Developers on GitHub

AI Agents

The stories that change how you build — and what you build with.

AI Agents Are About to Out-Consume Developers on GitHub — and That’s a Warning for Every Builder

What happened: GitHub COO Kyle Daigle dropped the stat that should reframe how you think about the agent era: developers and AI agents made a combined 7.38 billion commits on GitHub in September alone, and Daigle now says agents “could become bigger consumers of GitHub services than developers.” His prescription is a full re-think — scaling APIs for agents separately from humans, building agent-specific APIs, and rewriting documentation and tooling “so agents can consume GitHub efficiently.”

Why it matters: This isn’t a puff quote from a keynote. It’s a candid admission that agents are straining GitHub’s infrastructure hard enough to force an architecture change — including an August outage that lasted nearly eight hours and a painful migration from GitHub’s own datacenters to Azure. The deeper signal: if GitHub is caught off-guard by agent traffic, imagine what it’s doing to your own APIs and internal tools. Agents don’t behave like human users. They hammer endpoints, retry aggressively, and consume every scrap of documentation and permission surface you expose. Treating “agent access” as the same workload as “human access” is now a recognized failure mode.

What’s next: Daigle’s framing — “agent access as a fundamentally different workload from human access” — is the phrase you should steal for your own architecture reviews. If you’re building or orchestrating agents, start asking: do my APIs have rate limits and contracts designed for a machine that never gets tired? Is my documentation parseable by a model, not just readable by a person? The builders who answer those questions now are the ones whose agents won’t melt their own infrastructure later.

Source: GeekWire | Moneycontrol

Quick hit — OpenAI’s Frontier lands its first energy customer. ADNOC signed a multi-year deal with OpenAI and G42 to run agents across its oil-and-gas operations, becoming the first energy company on OpenAI’s Frontier agent platform. First use case: field-development planning that could compress months of work into hours. It’s a concrete look at how “sovereign infrastructure + frontier models + industry data” actually gets packaged for a giant enterprise.

Source: The Next Web

Quick hit — An agentic EMR with real revenue numbers. Physitrack opened its AI-agent EMR to clinics worldwide after an 800-clinic, 60-country waitlist, citing €2.6M–4M in projected ARR from 2027. With 70%+ activation among early users and a #3 G2 ranking in five weeks, it’s a rare vertical-agent story backed by actual traction, not a press-release promise.

Source: Yahoo Finance

AI News

What’s moving across the broader AI landscape.

Attackers are now running AI agents against enterprise networks. A China-linked group (Storm-0940) used ARTEX AI agents to automate Kerberoasting, password spraying, and lateral movement in a ConnectWise supply-chain attack — compressing exfiltration “from days to minutes” per Microsoft’s latest report. The uncomfortable takeaway for builders: the same agent autonomy you’re deploying for work is being deployed against you, and lateral movement now takes as little as four minutes.

Source: Aviatrix

Apple quietly acqui-hires an AI data startup. Apple took the talent, patents, and a non-exclusive license from Ascend.io, an AI data-engineering startup, in a deal disclosed via an EU filing — with Ascend posting a shutdown notice and no public Apple roadmap. It’s the latest quiet move in Apple’s pattern of talent-and-IP deals rather than splashy acquisitions.

Source: Entrepreneur India

OpenAI tightens ChatGPT’s violence-related controls. OpenAI is adding further restrictions on weapons-related information, including for teen accounts “even when a question appears harmless,” alongside broader limits for all users. It’s a reminder that model safety defaults keep shifting under your feet — worth tracking if your agent relies on a consumer-facing model’s current behavior.

Source: The Verge