Welcome back! Lines of code, job titles and internal tools are all under review at once. One wave of automation is testing how much software still belongs to humans, another stacks real professions into an exposure list, and a third sinks AI into the pipes that move tickets, approvals, and project notes around. The mood is part opportunity, part vertigo: intelligence spreading through the stack, from IDE to workflow to search box, and forcing people to ask what kind of work they actually want left on their side of the screen.
In today’s Generative AI Newsletter:
Anthropic Claude Code spooks software stocks.
Microsoft ranks 40 jobs as AI-exposed.
OpenAI voice agents power ServiceNow workflows.
Beloga lets workers aim AI at sources.
Latest Developments
Anthropic CEO Dario Amodei recently projected that models will likely perform nearly all software engineering tasks within twelve months. This forecast follows the viral release of Claude Code, a tool that allows researchers to manage entire codebases as unified reasoning problems rather than disconnected files. This surge in automated power caused an immediate drop in market value for firms relying on traditional human-led development. The engineer's role is moving from active building to high-level oversight of the entire system.
The impact of automated coding:
Production Gains: Early users report 5x productivity increases which led one CEO to cancel plans for new engineering hires.
Market Decline: The Morgan Stanley SaaS index dropped 15% this year as major players like Salesforce saw double-digit stock losses.
Engineering Velocity: Vercel CTO Malte Ubl completed a dormant year-long project in seven days using autonomous editing capabilities.
Automation Timeline: Dario Amodei claims that models will perform nearly all technical engineering duties before the end of this year.
As AI takes over coding, software value is no longer linked to human hours. This price crash shows the market does not see human code as special anymore. When a tool can copy a huge platform for a tiny price, old ways of selling software completely stop working. However, these fast goals assume AI can fix rare bugs and handle long-term care like humans do. Predicting a job will end is a tech story that misses how messy and complicated real life really is.
Microsoft researchers recently analyzed 200,000 real world conversations to map the intersection between software and human labor. Their findings rank 40 professions where automated logic mirrors daily tasks across classrooms and corporate offices. This study highlights a new vulnerability for Gen Z graduates and degree holding professionals who previously viewed their roles as safe from digital competition. The data forces a hard look at how knowledge workers manage their daily output in an era of rapid technical growth.
How the labor market reacts:
Top Exposure: Translators and historians rank as the most vulnerable roles because of high linguistic overlap with automated models.
Education Risk: Teachers and librarians face unexpected competition as software becomes better at explaining complex economic and business concepts.
Market Impact: Sales and customer service roles representing five million American workers must now compete with machine information systems.
Safe Sectors: Hands on occupations like dredge operators and water treatment specialists remain untouched due to their physical equipment requirements.
The release of this list creates a difficult environment for employees who are already managing the stress of a changing economy. While the study claims these tools will support workers rather than replace them, the reality of corporate budgeting often prioritizes the efficiency of software over the stability of a payroll. Using data from a company’s own users to predict the decline of their professions feels like a cold way to measure human value. In the end, no algorithm can replicate the personal mentorship or ethical judgment that remains the true heart of a professional career.
ServiceNow and OpenAI announced a multi-year deal to put OpenAI models inside ServiceNow’s business workflows so the system can take actions, not just answer questions. ServiceNow says it runs “more than 80 billion workflows every year.” That number matters because if AI sits inside the workflow layer, it can touch approvals, tickets, customer service and IT changes at scale. Controlling the operator layer allows for the taxation of work and attribution of blame when the system operates incorrectly due to breakage.
Here's what the deal holds:
Term: OpenAI becomes a preferred intelligence option inside ServiceNow for a three-year deal.
Control: ServiceNow utilizes the AI Control Tower to oversee and assess the actions permitted for the AI.
Market: The news reportedly pushed ServiceNow shares up over 2% premarket.
Product: Speech-to-speech agents that open and route cases and 'computer use' automation for old systems.
This seems like the Salesforce strategy where vendors want agents that can act inside business software. On the positive side, the system offers quicker issue resolutions, reduces the need for manual interactions, and introduces voice-driven support services. The downside is that an agent can approve the wrong change with perfect confidence and governance tools only help if companies actually use them. AI vendors seek widespread adoption, software giants prioritize innovation and customers raise concerns about accountability and ownership in the event of errors.
Beloga is a work search tool where you point your question at a specific source using simple @mentions like @Web, @Links, @Notes and @Files. It’s useful when you don’t want a generic answer, you want the answer from your own material or from one chosen place.
Core functions (and how to use them):
Source pick: Start with “@Web” or “@Notes” so the answer stays inside that source instead of mixing everything.
Link brief: Paste a handful of URLs, then ask “@Links Summarize into 5 bullets, then list 3 decisions” to turn reading into action.
Spreadsheet extraction: Upload a PDF, then ask “@Files Pull every number into a table with Metric | Value | Unit | Page," and paste it into Sheets.
Notes recall: Ask “@Notes Find where we decided X. Quote the exact line and include the note title" so you can verify quickly.
Shareable research: Share the same query thread with a teammate so you’re both looking at identical sources and outputs.
Try this yourself:
Open Beloga and paste 3 links you’ve been meaning to read (a doc, a blog post, a thread). Ask: “@Links Summarize these into 5 bullets and a to-do list.” Follow up with: “@Links What’s the one decision I should make from this?” Copy the to-do list into your notes and do the first item in 10 minutes.
