Just a general observation to share today. There’s a growing gap between how AI layoffs are being measured and how they’re being perceived in the market.
We’ve been living with LLMs in a meaningful way for a little over three years now, and the data tells conflicting stories.
Take an article I shared last week, where OpenAI’s Sam Altman suggested companies are “AI-Washing” layoffs; that they’re blaming AI for regular layoffs they would have done regardless.
Then just this week, Anthropic launched an “AI job disruption detector” that measures which occupations are most exposed to AI automation. Thus far, researchers found “limited evidence” that AI has meaningfully affected joblessness.
On the flip side, you have AI layoff trackers ike this one from flashpass, who today is reporting 79.3k people affected by AI-related layoffs.
And of course news last week of Jack Dorsey’s Block layoffs, purportedly motivated by AI, which cut 4k+ workers (nearly half of the company).
So what’s actually happening? The research suggests AI’s current impact is overstated, but the accelerating pace of these announcements suggests the disruption may be just beginning. Either way, we need to understand which narrative is closer to reality, because the answer shapes how we should be preparing our clients and our own teams for what’s next.
-Sage Levene, VP of Marketing, Hypha HubSpot Development
Open Mic
The Best Websites Are Built to Be Used
By John Patti, Senior Web Designer, Hypha HubSpot Development
Most conversations about great websites focus on the design. But from a development perspective, the real measure of a site comes after launch, when the client begins working with it day to day.
A website is really a tool that a marketing team lives inside. They’re using it to publish content, launch campaign pages, and make updates without pulling a designer in. When it’s built that way, it shows.
That’s where the behind-the-scenes choices start to matter. In HubSpot especially, those choices are everything.
Which fields are editable, how modules are named, whether a layout can stretch to fit a longer headline. Those choices quietly shape how much a team can actually do with the site once it’s live. A section that looks simple in a mockup can require a lot of thought behind the scenes to actually work that way.
Giving clients total control isn’t the goal either. Hand someone a blank canvas and the design falls apart fast. The real challenge is building guardrails that don’t feel like guardrails, page structures and modules that keep the brand intact while still letting a marketing team move without waiting on a developer.
When that balance is done well, the website becomes more than a finished project. It becomes a system. The marketing team isn’t rebuilding every new page from zero. They don’t need to pull in a designer every time something needs to change.
It’s easy to measure the success of a website by how it looks when launched but the real test comes after. When the site is still easy to use, easy to expand, and still doing the job it was built for.
A super comprehensive piece from Buffer that will help you tune up your social strategy:
“Across six platforms and nearly two million posts, accounts that reply to comments consistently outperform those that don’t — by as much as 42% on Threads and 30% on LinkedIn. That doesn’t mean replies cause engagement. But it’s one of the strongest patterns we found, and — we think — one of the most untapped.”
A disturbing facet of our digital reality. Even the most tech-savvy individual can fall for a scam, so it’s important to stay vigilant and show discretion when it comes to interacting with ads.
“Advertising accounted for more than 60% of the malware and phishing campaigns observed by The Media Trust in 2025. Instances of malware delivered via programmatic channels grew 45% year-on-year, per the report.
“The Media Trust said the threat has been compounded by advances in the $791 billion digital advertising space. Artificial intelligence makes it easier than ever to generate ads — such as celebrity deepfakes — and precision-target the most vulnerable consumers.”
Hypha Highlights
The ERP integration problem refers to the fundamental architectural mismatch that occurs when manufacturers attempt to use their ERP system’s built-in CRM module to manage customer relationships.
ERP systems are designed around transactional data—recording what happened, when, and at what cost. CRM systems are designed around relational data—tracking who said what, when, and what it means for future revenue.
These are different data models, and forcing one system to do both creates structural gaps that surface as daily operational friction: quotes that don’t reflect real customer history, sales reps working around the system, and management reports that are technically accurate but operationally useless.
Overwhelmed by the fast updates HubSpot makes to the platform? Make sure to subscribe to the weekly Product Update email so you don’t fall behind on releases.
In your Settings, go to Account Defaults > Feature Releases. Then toggle on the checkbox for weekly emails!
AI in Action
News, updates and tools from the AI industry.
It’s been a busy week:
Cursor introduced Cursor Automations, always-on AI agents that run on schedules or are triggered by events like Slack messages, Linear issues, or GitHub PRs to automate software development tasks. The agents are being used at Cursor for security reviews, incident response, test coverage, and bug triage, with external teams like Rippling using them to handle meeting summaries, Jira issue creation, and on-call handoffs.
U.S. uninstalls of ChatGPT’s mobile app surged 295% after OpenAI announced a partnership with the Department of Defense, while competitor Anthropic’s Claude saw downloads jump 37–51% and hit No. 1 on the U.S. iOS App Store after declining to work with the defense department over concerns about AI surveillance and autonomous weaponry. ChatGPT also saw 1-star reviews surge 775% while 5-star reviews dropped 50%, with Claude’s daily U.S. downloads surpassing ChatGPT’s for the first time.
Additionally, OpenAI launched GPT-5.4, its latest AI model featuring native computer use capabilities that allow it to operate computers and complete tasks across different applications by writing code, issuing keyboard and mouse commands, and using web browsers, with the company claiming it’s its “most factual yet.” Separately, OpenAI is developing a new code-hosting platform to rival Microsoft’s GitHub after engineers encountered frequent service disruptions, marking a bold move to compete directly against Microsoft despite the tech giant holding a significant stake in the firm.
The Pentagon formally notified Anthropic that it’s deemed the AI company and its products a supply-chain risk after contract negotiations broke down over Anthropic’s demands for assurances that its AI wouldn’t be used for mass surveillance of Americans or autonomous weapons deployment.
Google, Microsoft, Meta, Amazon, Oracle, xAI, and OpenAI signed a “Ratepayer Protection Pledge” at the White House committing to bear the cost of new electricity generation for their data centers to prevent Big Tech from driving up U.S. electricity costs for homes and small businesses. The agreement requires tech companies to bring or buy electricity supplies from new or expanded power plants, pay for power delivery system upgrades, and enter special rate agreements with utilities, though critics question whether it will get new electricity supplies built quickly enough to ease pressure on grids.
The Future of Life Institute released the Pro-Human AI Declaration after a January meeting that brought together 90 political, community, and thought leaders from across the political spectrum. The declaration, which focuses on centering AI development on humanity while avoiding power concentration and preserving human agency, garnered overwhelming support in polling, and comes as FLI deliberately excluded Big Tech companies from the drafting process after observing corporate interests dominating previous AI conferences.
Wikipedia editors implemented new policies and restricted contributors from the Open Knowledge Association (OKA) after discovering that AI-translated articles added hallucinations and errors, including swapped sources, unsourced sentences, and paragraphs from completely unrelated material. OKA, which instructs translators to use AI tools like ChatGPT, Gemini, or Grok for translations, will now issue blocks to translators who receive four warnings within six months, though the organization is adding a second AI review step to check AI-generated translations for errors.
“Get the newsletter for people who make newsletters. Every other Wednesday, we’ll send you resources to help you grow your newsletter, engage your readers, and drive more revenue. Join 10,000+ readers and learn how to build the right strategy for your newsletter!”
How can we help you?
Case Study: Building Content Infrastructure
A mid-market freight forwarder had no brand presence and no organic visibility in a search landscape dominated by global carriers with full content teams and far larger budgets. Their site looked like every other mid-market logistics company (functional but forgettable) and search engines had little reason to treat them as an authority.
We built a quarterly industry report based on live freight data, interviews with logistics operators, and analysis procurement teams actually use. Each report became five assets: the current edition gated for leads, the previous edition ungated for SEO, plus email campaigns, LinkedIn articles, and blog posts for each channel. We also rebuilt the site on HubSpot CMS with a full rebrand and a conversion-focused blog architecture.
The result: within 18 months they were ranking alongside global carriers on competitive search terms. Years later, content from that period still holds a position-two ranking for high-volume queries and drives thousands of organic clicks each month. One strong quarterly publication, produced consistently and repurposed across channels, created more lasting value than a higher volume of lighter content.
Publishing consistently but not seeing organic traction? Contact Hypha to build content infrastructure that compounds, not just content that disappears.
