I keep a running list of things I come across. Articles, headlines, quiet announcements, papers I bookmark and tell myself I'll revisit. Most of the time it stays in my head and gets lost.
But April and May 2026 have been wild. So I sat down and went through the last forty-five days across the things I care about. Tech, AI, medicine, and senior care. Twenty stories made the cut. None of these are speculation. All of them happened.
This is mostly for me, honestly. A way to slow down and actually pay attention to what is happening in the world right now. But if any of it is useful to you too, I'm glad.
AI: What Shipped, Who Is Winning
1. Anthropic found thousands of zero-day bugs and locked the model away
Anthropic announced Claude Mythos on April 7. They turned a frontier model loose on the world's software and it autonomously found thousands of high-severity vulnerabilities, including a twenty-seven-year-old bug in OpenBSD, a sixteen-year-old flaw in FFmpeg, and a remote-code-execution hole in FreeBSD's NFS server that it reached by chaining six requests together. Over ninety-nine percent of what it found is still unpatched.
The twist: they refused to release the model publicly. They locked it down to about fifty partners under something called Project Glasswing, with a dozen named at launch.. Apple, Google, Amazon Web Services, JPMorgan, Microsoft, NVIDIA, Cisco, Broadcom, CrowdStrike, Palo Alto Networks and the Linux Foundation. First time I can remember a major lab saying out loud that their best model was too dangerous to ship.
I've been a Claude user for two and a half years now. I'm on the Max plan. Opus 4.7 is already mind-blowing to me, and Mythos is apparently another step up from that. I don't know if Anthropic is being honest about the danger or if some of this is marketing. But the amount of stuff Anthropic ships is just mind-blowing to me. They keep moving.
2. OpenAI shipped GPT-5.5 with a million-token context window at five dollars
GPT-5.5 dropped April 23 with a one-million-token context window at five dollars per million input tokens. Three days later they made it the default for free ChatGPT users. Long-context work that cost fifty dollars a year ago now costs a dollar.
I think OpenAI needed to do this. They were losing ground to Anthropic and Google, and ChatGPT is still where most of the general population lives. If they had not beefed up their free and basic tiers, they would have hemorrhaged regular users to Gemini and Claude. Good call. Even Sam Altman can't let the casual user drift.
3. Anthropic in talks at a $900 billion valuation. Google triples down.
By mid-May, Anthropic was in talks for a $30 billion round at a $900 billion valuation. More than double the $380 billion it closed at in February. Google separately confirmed a $10 billion investment with another $30 billion contingent on milestones.
This is almost a death sentence to OpenAI. Almost. But I hate to bet against Sam Altman. He may be the guy everyone is targeting, but their current ChatGPT model and Codex are shipping just as well as Anthropic, and even better on some benchmarks I've seen. So it depends on who can grind faster and longer. But Google as a backer for your competitor? That's not great for Sam.
4. Anthropic launched Claude for Small Business
May 13: Anthropic shipped a packaged product aimed straight at small operators. Fifteen prebuilt agentic workflows wired into QuickBooks, HubSpot, Canva, DocuSign, PayPal, Google Workspace, and Microsoft 365. Their head of small business explicitly named the target: "the fifteen-person HVAC company, the thirty-person landscaper, the fifty-person real estate brokerage."
I love this. I've been using Claude almost exclusively for two and a half years. I started with ChatGPT, moved over, and I've used Claude to build internal tools for Golden Pines and for other small businesses. The fact that Anthropic is now meeting small operators where they are, packaged workflows, connectors, a free training tour, is great for adoption. It lowers the bar and the cost for everyone.
5. A two-person company hit a $1.8 billion run-rate
A guy named Matthew Gallagher started a telehealth GLP-1 company called Medvi from his apartment in September 2024. Twenty thousand dollars. Him and his brother Elliot. Year one: $401 million in revenue, sixteen percent net margin. 2026 is tracking toward a $1.8 billion run-rate. ChatGPT, Claude, and Grok wrote most of the platform. For comparison, Hims and Hers does similar work with 2,442 employees at 5.5 percent margin.
One thing I want to put next to that, because it belongs there.. the FDA sent Medvi a warning letter in February over misleading product claims, and the complaints have kept coming since. Fast is not the same as sound.
Still, I think this is where we're all going. Leveraging AI to expand reach with fewer people. It's a disruption, I know. My hope is that for every job lost, new fields and new positions get created. That has happened every time in history. But when we're in the middle of it, it's hard to see far enough ahead to know what those new jobs will be.
6. Stanford says entry-level developer jobs are down twenty percent
The 2026 Stanford AI Index came out in April. Software developer employment for workers aged twenty-two to twenty-five is down nearly twenty percent since 2024. Entry-level call center hiring is down fifteen percent. Workers thirty and older in the same AI-exposed jobs grew their employment six to twelve percent. Erik Brynjolfsson at Stanford is calling entry-level cohorts "canaries in the coal mine."
This is the thing I was talking about. It's what I worry about when I think about my friends' kids, my own family. The displacement is real. But I keep coming back to history. The internet boom did this. The mechanization of agriculture did this. Old jobs went away. New jobs got created. We're in the middle of it, so we can't see clearly. The new jobs are being invented right now. We just don't know what they are yet.
Medical Breakthroughs Worth Knowing
7. The first non-antipsychotic for dementia agitation got approved
April 30: the FDA approved Auvelity (dextromethorphan-bupropion) for agitation associated with Alzheimer's. It is the first non-antipsychotic approved for this. In the pivotal trial it cut relapse from 28.6 percent on placebo to 8.4 percent on the drug, and no boxed mortality warning, which every antipsychotic in dementia care carries.
This matters to me directly. At Golden Pines, we use a layered approach to agitation. Cognitive behavioral techniques. Prescribed antipsychotics when clinically appropriate. Staff training that's honestly the biggest variable of all. Another tool, especially one without the mortality warning, is greatly appreciated. Agitation drives roughly half of all dementia-related emergency room visits. Advancing care in this field is important. The residents deserve every option we can give them.
8. The first oral GLP-1 pill got approved
April 1: the FDA approved Lilly's Foundayo (orforglipron). The first oral, non-peptide GLP-1 pill. Twelve-point-four percent body weight reduction at the highest dose. No needles, no refrigeration, no fasting window.
I'm still uncertain about GLP-1s, even in the injectable form. They hit the market so suddenly. Massive traction in such a short time, and the long-term picture is still being written. But if a pill gives people more options, especially older adults who can't deal with injections, I'm all for it. I just hope it's not horrendously expensive. That's always the question with breakthrough drugs.
9. The first gene therapy for genetic hearing loss got approved
April 23: the FDA approved Otarmeni for children and adults with profound hearing loss from mutations in the OTOF gene. A single surgical injection into the cochlea. Eighty percent of pediatric patients in the trial regained measurable hearing. Regeneron is providing it free in the US.
Hearing loss, by the way, is the single largest modifiable risk factor for dementia according to the Lancet Commission. So if you can fix hearing, you also chip away at dementia risk down the road. Advances in health technology like this have a ripple effect. The first one of anything is the hardest. Once you prove you can deliver gene therapy into the inner ear, you open the door to all sorts of things, including age-related hearing loss, which is the bigger prize.
10. CRISPR cleared Phase 3 inside living humans
April 27: Intellia's CRISPR therapy for hereditary angioedema cleared Phase 3. One IV infusion. Sixty-two percent of patients were entirely attack-free for six months versus eleven percent on placebo. No serious adverse events. It is the first in vivo CRISPR gene-editing therapy ever to make it through Phase 3.
This is the one. I can't believe it isn't getting more attention. This kind of advancement, where you could potentially cure or treat so many ailments with a single infusion that edits a gene inside a living person, is hard to overstate. Cholesterol. ApoE. Amyloid pathways for Alzheimer's. The door just got kicked open for all of it. I couldn't be more excited about this one. Put it on your list.
11. A pancreatic cancer vaccine is keeping people alive at six years
At AACR in April, Memorial Sloan Kettering presented six-year follow-up on a personalized mRNA pancreatic cancer vaccine. Sixteen post-surgery patients. Eight of them mounted strong T-cell responses to the vaccine. Of those eight, seven are still alive six years later.
Pancreatic cancer has a twelve percent five-year survival rate. It is essentially a death sentence. So seven out of eight at six years is staggering.
I have to be honest, this one hit me. I love that mRNA technology is being used not just for COVID. I thought about this during the height of the pandemic, actually. If they could spin up an mRNA vaccine for COVID in months, imagine what that technology could do for things like cancer. And here we're. Pancreatic cancer. The most devastating one. The one that took Patrick Swayze and Alex Trebek. Watching this technology finally show what it can really do, years out from a death sentence, is amazing to me.
Senior Care, My Industry
12. Aline and PointClickCare both baked AI into senior living software
April 23, Aline announced Aline Intelligence. AI embedded across their CRM, care platform, menus, and referral systems. Hospital intake decisions compressed from thirty minutes to under four. A few weeks later, PointClickCare brought Chart Advisor to senior living, a tool it had already been running in skilled nursing that reads chart entries in real time and flags fall and medication risks at the point of charting.
These are the two big platforms a lot of senior living operators run on. So AI is arriving in the back office whether operators ask for it or not. I think this is great news. They have to adapt. Otherwise their subscription rates will drop drastically as small operators, and even big operators, realize they can customize similar tools on their own. I've already built internal tools at Golden Pines using Claude. If I can do it, so can anyone.
13. Senior housing occupancy hit 89.5 percent, nineteen straight quarters of gains
NIC released their Q1 2026 numbers in April. Occupancy is at 89.5 percent, up from 89.1 at the end of last year. Independent living is over ninety-one. New construction is at a thirteen-year low. Inventory grew only 0.4 percent year over year, a record low. Boston is at 93.6 percent. Atlanta is the softest at 86.
This is the tip of the iceberg. We're at a tipping point in senior care. The Baby Boomers are aging into care faster than we're building capacity, and the math doesn't work. We need lawmakers, business leaders, and senior advocates working together, soon, to make senior living and senior care more affordable. Otherwise seniors get left on the street, or nursing homes get overrun and become critical. I'm probably going to write a longer piece on this. The crisis is here. We're just not naming it yet.
14. Michigan came through for the senior care industry
Michigan's FY 2026 budget included a $621,000 legislatively directed grant to the Michigan Assisted Living Association, MALA, for online training built on their learning management system. The enacted budget also put $700,000 into oversight of state-licensed homes for the aged and $428,500 into nursing home complaint intake and investigations. House Bill 5466 would restrict the state health department's authority to limit visitation during emergencies, a direct response to COVID-era policies that hurt residents and families.
I'm a big supporter of MALA. Good news for all of us in senior care in Michigan. The training is free or subsidized for members, and that kind of help matters a lot for small operators who don't have a corporate training department. The visitation bill is also overdue. We learned the hard way during COVID that locking residents away from family is its own kind of harm.
15. Elder fraud hit $7.7 billion. Voice-cloning is the new weapon.
AARP flagged the federal numbers in April: Americans sixty and older reported $7.7 billion in fraud losses last year, according to the FBI's Internet Crime Complaint Center. A sixty percent jump from 2024. The FBI's Internet Crime Report, the first one in twenty-five years with a standalone AI category, counted $893 million in AI-related fraud losses across 22,364 complaints, and about 39 percent of that landed on people sixty and over. Three seconds of audio scraped from social media is enough to clone a grandchild's voice.
It's a sad reality. There were already high incidences of elder fraud before AI. Now it's ramping up to a level that I think will catch a lot of families off guard. My expectation, and my hope, is that technology will start using AI to fight this too. Fire with fire. Better spam detection. Better authentication. Better awareness at the family level. We have started talking to our resident families about safe words. The kind of word a real grandkid would know and a voice clone wouldn't. Costs nothing. Worth doing.
16. The Big Beautiful Bill Medicaid cuts are starting to bite
The One Big Beautiful Bill, signed last July, is cutting more than a trillion dollars from federal health programs over time. By April, more than 800 hospitals, nursing homes, maternity wards, and psychiatric facilities had already closed, cut services, or were at risk. Idaho enacted a four percent across-the-board Medicaid provider rate reduction. Colorado reversed a planned increase. About sixty-three percent of nursing home residents nationally rely on Medicaid.
This one is actually terrifying. Many seniors are already hard up. The fact that it's now going to be harder to get seniors without funding into a proper facility. That's what we have to stop. The senior home crisis has to be something all sectors focus on. Cutting funding isn't a solution. Without clear alternative solutions, we risk the health and welfare of our seniors even more. I don't have an answer here. I just know that the people who do have answers aren't the ones cutting the budget.
17. The AI for Main Street Act cleared the House 395 to 14
The House passed the AI for Main Street Act in January by 395 to 14, and it is now with the Senate, where a companion bill already has sponsors from both parties. Almost unanimous, which is rare for anything these days. It makes AI adoption a formal SBA priority, funds AI literacy training through 900+ Small Business Development Center locations, SCORE chapters, Women's Business Centers, and Veterans Business Outreach Centers nationwide. Grants are coming in the second half of 2026.
I love this. I'm going to dig in and see if Golden Pines, or some of my other small business ventures, can take advantage. I might write a separate piece on the details once I understand the actual mechanics. For a small operator who is already using AI to run his business, federal training money and grant programs landing in the local Michigan SBDC office is a real thing.
The Bigger Picture
18. A Chinese company built a battery that charges in six minutes
April 21: CATL, the Chinese battery maker that supplies roughly a third of the world's EVs, unveiled a battery that charges from ten to ninety-eight percent in six minutes and twenty-seven seconds. Same day, they showed a sedan-class pack with a 1,500 kilometer range. That is about 930 miles on one charge.
This is what I want to see in electric vehicles. The speed of charging and the length of the drive. Those have always been my two reasons for not buying a Tesla yet. If both of those were in a Tesla, or any EV I could buy in Michigan, I'd already have one. The two arguments against EVs collapsed on the same day, from the supplier that makes batteries for a third of the world's EVs. That's the kind of step-change that actually re-prices an industry.
19. Commonwealth Fusion filed for a power grid connection
On April 28, Commonwealth Fusion Systems filed the first-ever grid interconnection request from a fusion company, to PJM, the grid operator for thirteen states. They want to connect their planned fusion plant in Chesterfield County, Virginia. Their main reactor in Devens, Massachusetts is seventy-five percent built.
For seventy years, the joke about fusion was that it's thirty years away, and always will be. A grid interconnection request is the boring paperwork a real power plant files. No fusion company has ever filed one before. So this is the moment fusion moves from a physics question to an engineering and permitting question.
If this works, it is huge for the environment. We would lessen our dependency on foreign oil. We would accelerate EV adoption. We would replace fossil fuels in our daily lives in ways that have only been theoretical. CFS is private, by the way. Backed by Bill Gates and Google and Khosla. Helion is private too. There is no easy way for a regular person to invest in pure-play fusion yet. But the day one of these companies goes public, that will be a moment to watch.
20. Quantum computers got a lot closer, and so did losing every secret we have
In early April, Caltech physicists and a startup called Oratomic published an error-correction scheme suggesting a quantum computer capable of breaking RSA-2048 encryption, the math behind almost everything secure on the internet, needs something like ten to twenty thousand qubits, not the millions everyone assumed. Harvard separately published a fault-tolerance advance that pulls large-scale quantum timelines forward by five to ten years.
The scariest part is not that today's encryption breaks tomorrow. It is that bad actors are already harvesting massive amounts of encrypted traffic right now, banking data, government cables, private medical records, and storing it. They cannot read it yet. But when a quantum computer comes online, they decrypt all of it retroactively. Everything we are sending in 2026 becomes readable somewhere around 2032.
Damn. I don't really know what to say with this one. Bank data and passwords we can change. But people's journals. Diaries. Private messages they thought no one would ever see. All of that comes out the other side of the decade. Oh boy.
A Note to Myself
This is what twenty stories looks like compressed into one sitting. I came in to read the news and I came out with three or four ideas for future articles. One on the senior housing crisis, one on what mRNA technology might do for the rest of medicine, one on the AI for Main Street Act, and probably a fourth on what investing in this stuff looks like.
The world is moving fast. Faster than I can keep up with most weeks. But sitting with twenty things at once, in one place, helps me see the shape of it. AI is shipping. Medicine is finally hitting some of the big swings it has been promising for decades. Senior care, my own industry, is at a tipping point. And the bigger picture, energy, encryption, the structure of the economy, is shifting in ways that will outlast any one news cycle.
I'll keep doing this. Probably twice a month from here on out. Mostly for me. But if you're reading and any of this was useful to you, then I'm glad.
