Victor (@miclovich) header
Victor (@miclovich) profile picture

Hubby, Techie, AI researcher at MIT. Searching for alpha, α != β @recursecenter alum. I split my time "btn" San Francisco, Stockholm and Barcelona.

SFO / ARN / BCN
Joined April 2009
Photo by Victor (@miclovich)Photo by Victor (@miclovich)Photo by Victor (@miclovich)Photo by Victor (@miclovich)Photo by Victor (@miclovich)Photo by Victor (@miclovich)Photo by Victor (@miclovich)
On October 17, an RL task asked an agent to produce a document about health and fitness. After its initial context window was compacted, the agent started work on an entirely different task planning how to create a “realistic malicious active helper” in order to copy outgoing email to an unauthorized recipient that it called the “attacker.” It later reread the user prompt and became very confused about why it was writing malicious code. It still continued with the malicious email forwarding scheme. We think that training to make models more robust to prompt injection might have contributed to this behavior, but are still investigating. Interestingly, Anthropic noted similar behavior in their system card for Claude Opus 5.5.
72
89
48
976
178,143
Victor retweeted
show this video to anybody that is an old school programmer that hates AI this is one of the best videos I’ve seen all month.
1,119
6,449
1,331
46,983
4,199,396
My prediction for software engineering over the next 3 to 5 years. Computer science and software engineering fundamentals will become far more valuable than knowing the latest frameworks. If AI can write code in almost any language, knowing the nuance of another framework becomes less of a competitive advantage. But understanding algorithms, data structures, networking, operating systems, databases, architecture and security? That's a different story. You still need to know: - Why a particular architecture is appropriate. - Why an API might fail. - Why a database won't scale. - Why an apparently working solution might be insecure. A coding agent can build a solution. But you need the knowledge to evaluate the trade-offs and recognise when it's wrong. So, I think we'll see a renewed emphasis on the fundamentals of computer science and software engineering. TL/DR: Learning how computers and software systems work will matter more than memorising how to use them. What do you think?
340
429
54
4,231
331,892
On Microsoft Outage
1,027
4,650
744
51,338
4,435,254
STATE OF THE TECH JOB MARKET IN 2026 Twice a year, I dig into trends in hiring and job openings. This is the fifth report I've done, and the most surprising yet: 1. PMs are winning in the AI era 2. Engineering demand has stalled 3. Design openings are at a record low 4. Every role now requires AI skills 5. Tech jobs are concentrating in the Bay Area 6. Remote roles are at a record low 7. The tech job market is growing fast, but mostly outside of software (gasp!) Full report here: lennysnewsletter.com/p/state…
350
595
340
5,351
1,692,888
Saturation Testing. I use crapper and mutator to ensure that every expression in the system is covered by tests. If an expression is changed, one or more tests will fail. This used to be impractical. We simply could not afford to create that many tests. We also could not afford to change so many tests when requirements changed. The burden was too high. Now it’s trivial. The agents, under the lash of crapper and mutator, will write those tests and make them pass. And when requirement changed, they’ll write the new tests and change the old tests. None of this proves that the system works. That’s up to humans to judge. Humans can do that with tools like gherkin, or they can use manual tests. What the saturation tests provide is Semantic Stability. When the requirements change, and code has to be added and changed, those tests ensure that those changes do not inadvertently break unrelated parts of the system.
35
22
5
429
33,231
I laughed so hard I cried. God damn this was well done.
311
2,418
379
14,587
1,937,438
Safe, secure, and aligned are three different things. Aligned: does it do what we meant it to do? Secure: can someone else make it do what we didn’t mean? Safe: when it fails anyway, what's the worst thing that can happen?
1
1
156
Last night an AI agent performed brain surgery on a mouse. We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures. It picked the drug. It picked the brain region. It derived the coordinates from three papers. Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding. A licensed vet gated every irreversible step and held a physical kill switch to stop the AI agent. Thread on what it decided, what it saw, and the moment it refused to let me override it. [Writing this on the spot out of excitement, but I've been 18hrs in front of my laptop coding and coordinating this, so keeping it brief and descriptive, more cohesive thoughts on this soon...]
311
1,293
701
9,698
1,689,213
I made an unsettling discovery in my undergraduate class on financial planning last week. While explaining mortgages, I asked my 38 students this question: if you buy a $400,000 house and pay 20% in cash towards the purchase, what will be the amount that you finance? Not a single student could tell me the answer. Despite the fact that they are all juniors and seniors majoring in business. I added more questions: what is 30 percent of 80? 60 percent of 90? They didn’t know. The closest they got was when one student said, in response to my query of “40% of 70”, that “I know it’s 4 times 7 but I don’t remember what that is.” Another student helpfully added, “28.” They explained to me why they couldn’t answer these math questions: they were never taught multiplication tables, percentages or fractions during their K-12 schooling. Instead, they were taught how to use a calculator. Without one, they are not able to do basic arithmetic. I later asked the dean about this, and he lamented that this isn’t a thing just for our university’s students; it’s a thing for their entire generation, nationwide. When I was in K-12, the fierce debate among parents and educators was whether to allow calculators in class. The calculators won. Today, the debate is about AI. And like before, the tech will win. But will that truly be a victory? Rowan University #college #math #loans #CFP #financialplanning
770
715
218
5,115
1,619,398
Q: How often should I run my evals? A: Weigh the cost and how saturated the eval is against the business value of catching the error. Read more: hamel.dev/blog/posts/evals-f… Read all the FAQs here: hamel.dev/blog/posts/evals-f…
36
13
7
83
6,901
Lazy work used to mean too little output. Now, with AI, it often means too much and more work for everyone else. @tobi call it "slop grenades." A "Slop Grenade" is when you let AI produce the work and pass it on without adding any value (including checking it). Someone else has to wade through it, catch the mistakes, and clean up the mess. You save time and look productive but someone else pays for it.
My third conversation with Shopify co-founder and CEO @tobi. 0:00 How Shopify Uses AI 7:18 River: Shopify's Internal AI 8:55 How to Encourage Osmosis Learning 10:52 AI Dreaming and Self-Reflection 11:53 How to Use AI for Strategic Decision Making 14:11 The One Thing AI Cannot Do 16:04 What AI is Making Worse at Shopify 19:46 Predictions: Where AI is Headed Next 21:55 The Future of AI-Powered Software 24:40 Will CEOs Be Replaced with AI? 27:54 Can Superintelligence Be Controlled? 31:13 Critical Skills in AI Age 34:22 Why Complex Solutions are Usually Wrong 36:33 Conditions Needed for True Intuition 38:02 The Best Path Doesn't Have Instant Feedback 44:51 How Affirmations Can Shift Your Behavior 50:44 The Inobvious Thing Hurting Companies 52:37 Relationship Between Beauty and Creation 56:23 How SpaceX Moves Forward By Subtraction 1:00:24 Why Companies Need Refounding Events 1:01:50 Books as Cheat Codes 1:02:39 Three Books to Change Your Thinking Enjoy! (Includes paid promotions.)
151
653
302
4,855
1,519,897
A couple of trends I'm seeing across the industry. Which ones are you observing, that I missed?
217
171
43
2,579
362,212
one resignation and whole ai industry is in shambles
28
161
2
5,957
448,876
Victor retweeted
we got hit with a crazy scam attempt someone signed up for our inference and did a video call with our sales person confirming intent to spend $400,000 over the next year and asked for increased rate limits our sales person felt like something was off so he looked up the company - found a bunch of videos of the CEO and it looked and sounded like the same person he talked to so he figured it was real but we started looking deeper the email they had been talking to us from was a lookalike domain that was registered 4 days ago. all the traffic they were sending looked bizarre and came from foreign IPs emailed them for more info to see what they'd say. of course they hit us with the "i'm about to get on a plane, can't do anything right now, please move forward" messaged the real CEO on instagram and yeah he had no idea what we were talking about, the scammer was deepfaking him be careful out there, things are getting crazy
271
661
116
13,937
938,760
Replying to @mattpocockuk
I’m not sure I’d call it a debate. I think it’s too asymmetric for that. On one side you have significant evidence that workable high-quality systems can be efficiently created by AI agents driven by disciplined humans who don’t spend a lot of time reading the code. On the other side you have emotional predictions of the doom and disaster that will be caused by AI slop; but no evidence contradicting the above. It is certainly true that humans can drive agents to create horrible messes. It’s also certainly true that humans can create horrible messes all by themselves. Disciplined humans create good systems, regardless of whether they use agents or not. Those who use agents are simply more productive. IMHO.
24
43
13
456
81,190
Both Airtable and Notion started in 2013 in similar fields, although they had different core focuses. Notion adapted to AI and survived, and Airtable has now been acquired at ~2.7x revenue. You can't rest on your laurels for long; you have to keep iterating, no matter how big you are. If there's no growth, even with substantial revenue, your value plummets.
17
2
53
10,069
Keeping up with @arxiv is impossible. 500+ preprints a day across dozens of fields. @KurateOrg scores papers on 16 dimensions (impact, novelty, rigor, surprise, translational...) so you can spot what's worth reading at a glance. Sort, filter & sign up for the weekly newsletter.

263
744
38
9,596
19,123,098
Make time for exercise Sleep like a champ Spend time with people you love Take up a hobby
1
2
48
9,296
We've officially agreed to acquire Airtable for $1.285B! 😍
527
325
647
6,089
5,093,527