I write code for fun and profit.
Bulgaria
Joined January 2010
- Tweets3.2K
- Following427
- Followers226
- Likes21.7K
Ivan Tanev retweeted
I'm looking for my next role 👀
TS SDKs, DX, agent tooling. remote/hybrid from Boston.
if you've enjoyed ArkType, Standard Schema or my TypeScript perf work, a repost would mean a ton 🙏
ssalbdivad.com
Just for context:
The French attack was because all fridges were exposed on the net, and the user/pass was admin/(current date).
Nothing to do with advanced cyber AI post-cryptography xD
I think it's plausible that within a year we'll see AI-based cyberattacks on the power grid, financial system, internet infra, or water. Things like the French supermarket cyberattack but worse. As basic prep I would recommend:
Cash: Get a few weeks of spending in cash
Internet: Get a Starlink Mini so that you have Internet if local fiber or cell is down (~$150 plus $10/month in standby mode)
Power, minimal: Enough to run Starlink Mini and phones (~$1,250)
Daily need: about 1 kWh (Starlink Mini 0.7, phones and laptops 0.2; can limit runtime)
Battery: 1 kWh LiFePO4, e.g. EcoFlow DELTA 3 Plus
Solar: 1x400W folding panel, which refills about 1 kWh on a sunny day
Power, better: Enough to run a fridge, Starlink Mini and phones (~$5,000)
Daily need: about 2.5 kWh (fridge 1.5, Starlink Mini 0.7, phones and laptops 0.2)
Battery: 4 kWh LiFePO4, e.g. EcoFlow DELTA Pro 3
Solar: 3x400W folding panels, which refill about 3 kWh/day in sun
Generator: ~2,500W dual-fuel inverter generator; refills battery in a few hours even if weather is bad; get a few 20 lb propane tanks
Water: 14+ gal of water per person for two weeks
Food: 2+ weeks of shelf-stable food
Comms: Hand-crank NOAA emergency radio
Ivan Tanev retweeted
Context pollution through code rot is the worst. You throw in a slop grenade, and your subsequent work is slightly worse and less efficient, and you need more and more resources to keep going.
Subagents don't solve this issue, they actually only make it worse long-term. Rigorously ensuring code quality however you can is an absolute must if you want to make it
Ivan Tanev retweeted
We’re in the “phone phreaking” era right now for AI agents. AI agents are selling you on convenience; how they will actually work at scale is very different: a cautionary tale
Back in the 60s and 70s, some kids figured out you could use toy whistles and the like to control AT&T’s phone network. AT&T actually published the relevant frequencies in open literature; you could just look them up. AT&T did this for the benefit of telephone engineers; it never considered that the system could be hacked.
The system worked so long as no one took advantage of it. It was convenient. But once knowledge spread of *how* it worked, and how easy it was to reverse engineer, it broke down. AT&T shifted over to a different system, and phreaking came to an end
AI agents are doing the exact same thing. They’re taking advantage of a system that works on human trust, and also on the friction - the time cost - that having to call or chat to a customer service agent imposes. Everyone has made the calculation about whether or not it was “worth it” to make a complaint or request or to raise an issue. Often, it isn’t. So you just don’t say anything.
AI agents reduce the cost of complaints/requests to ~ zero. Anything free is consumed at much higher rates, so we should expect the total volume of this category to explode.
Further, today companies and their employees typically assume that the complaints or requests they receive have an actual person behind them. Once AI agents become a mainstream thing, this will no longer be true. Now the assumption will be reversed: that any given customer interaction is actually coming from a bot.
Just as early car ownership was liberating, but mass ownership represents a constraint on your freedom, the same will be true with AI agents.
People using these agents today are getting good responses from companies because they are benefiting from the (now unwarranted) assumption that these requests are coming from a person. They’ll
no longer think this in the near future. The ways they train their customer service people to deal with customer requests will change, too. You don’t have to be nice to an agent. And they won’t be. The return to AI agents will decline as companies build entirely new workflows to deal with the huge volume of automated requests they’ll be getting from bots in the near future
The system today works on trust: that there’s a real person behind each customer interaction, and that if somebody has reached out to you, they’re concerned or mad or scared or whatever enough to do so. The contact is a signal that some minimal threshold of importance has been met. The business model is built on this belief about customer behavior.
But that trust is broken by ubiquitous AI agents. None of the current assumptions will hold true in the future. Once the customer population is mostly bots, the nature of customer service (at least online) is fundamentally changed
Agents will make the online environment even worse. I’m very bearish on the category. All the tech guys seem excited by this new frictionless world of automated convenience where the agents are getting you automatic refunds and canceling your old subscriptions and getting you that free upgrade on your next flight
Please.
You’ve got to think non-linearly. All these new frictions will need to be introduced to keep the system working smoothly. Agents rely on deception; that there’s a person behind them. But that’s not a safe assumption anywhere. So the system will adjust, just as it did with the phreakers: from default open and trusting, to adversarial: prove you’re human
I really can't fathom the not-reading-code crowd. I can see some code that's worth maybe less attention, and only if you have good lint/test on top of it - I probably don't care that much about UI only stuff. But I do care about a state machine driving the ui, the network level, the backend...
Vibe coding is fine - actually required nowadays - but at the end you should be able to answer how any customer facing production code works. You should understand the data structures, state transitions, performance characteristics and store format. It's non-negotiable.
Ivan Tanev retweeted
I am losing roughly $800 a month by working on MSW full-time while it's one of the most used projects in the ecosystem. Please, if MSW provides you any value, consider becoming a sponsor. I mean it when I say every contribution counts.
Thank you ❤️
github.com/sponsors/mswjs
Ivan Tanev retweeted
>be me
>discover effective altruism
>apparently normal charity is inefficient
>why donate to random sad thing when spreadsheet can tell you optimal sad thing
>fair enough
>buy mosquito nets
>save lives
>numbers look good
>feel powerful
>couple years later
>someone asks an innocent question
>why only count people alive today
>huh
>future people matter too
>obviously
>my grandchildren shouldn't matter less just because they haven't spawned yet
>reasonable.jpg
>keep following logic
>what about their grandchildren
>also yes
>what about people in 500 years
>sure
>5000 years
>why not
>500 million years
>starting to get weird but morality is morality
>open calculator
>humanity could survive for an astronomically long time
>could colonize galaxy
>could have trillions upon trillions of descendants
>maybe digital people too
>maybe simulated civilizations
>maybe dyson spheres full of happy uploaded minds
>calculator starts smoking
>realize currently living humans are rounding error
>8 billion people suddenly looking extremely beta
>future contains potentially 10^something people
>can't even fit beneficiaries in google sheets
>new moral priority unlocked
>protect the long-term future
>stop thinking in units of "people helped"
>start thinking in "fraction of cosmic endowment preserved"
>malaria?
>terrible
>but only kills existing humans
>AI extinction could delete the entire light cone
>nuclear war could permanently derail civilization
>bad institutions could lock in terrible values for ten million years
>someone invents wrong constitution in 2140
>quadrillions suffer
>better fund governance workshop now
>friend says maybe we should improve hospitals
>explain opportunity cost
>friend says hospitals are full of actual sick people
>explain scope sensitivity
>friend stops inviting me to dinner
>need to decide what to fund
>easy
>expected value
>suppose project has one in a million chance of preventing extinction
>sounds tiny
>but extinction destroys 10^50 future lives
>multiply
>mother of god
>$10 million project has expected value of several galaxies
>charity evaluation complete
>someone asks where the one-in-a-million number came from
>expert judgement
>which expert
>us
>how calibrated
>extremely thoughtfully
>reduce estimate to one in ten million to be conservative
>still beats curing cancer by 38 orders of magnitude
>epistemic robustness achieved
>someone says maybe project doesn't work
>assign 20% chance
>still astronomical
>maybe project makes problem worse
>assign 5% chance
>still astronomical
>why 5
>because 30 felt pessimistic
>publish 46-page report
>contains seventeen sensitivity analyses
>every sensitivity analysis begins after assuming intervention has positive sign
>critic says you're multiplying enormous hypothetical stakes by extremely uncertain probabilities
>yes
>that's literally why it's important
>critic says the uncertainty might be structural rather than numerical
>make probability smaller
>critic says no, I mean maybe your model is wrong
>make probability smaller again
>critic begins rubbing temples
>discover AI safety
>perfect longtermist cause
>AI might kill everyone
>or create utopia
>or seize galaxy
>or tile universe with paperclips
>or create billions of conscious software minds
>finally a problem with numbers big enough for me
>start AI safety nonprofit
>mission: prevent dangerous AI
>hire smartest people available
>smartest people immediately start building better AI to understand dangerous AI
>interesting
>we must understand capabilities to understand safety
>we must scale models to study alignment
>we must race ahead so less responsible actors don't get there first
>we must deploy systems to learn how deployment can go wrong
>we must build the thing quickly because building the thing quickly is dangerous
>outsider asks why the people most worried about AI apocalypse all work at AI companies
>complicated field
>company releases stronger model
>very concerned
>company begins training even stronger model
>extremely concerned
>company raises $14 billion
>concern reaches unprecedented levels
>need to influence government
>future is at stake
>normal democratic process too slow
>politicians don't understand exponential curves
>public doesn't understand x-risk
>experts must guide them
>who counts as expert
>people who understand x-risk
>who understands x-risk
>our friends
>someone objects that this seems politically convenient
>explain we're representing future generations
>future generations unavailable for comment
>develop concept of value lock-in
>terrifying possibility that one ideology controls civilization forever
>therefore extremely important that civilization adopts correct values before lock-in
>whose values
>let's circle back
>begin with impartial morality
>end with small group of people deciding what quadrillions of hypothetical beings would want
>beautiful arc
>meanwhile actual humans keep doing annoying things
>voting wrong
>having parochial attachments
>loving family more than strangers
>caring about local community
>getting upset when told their suffering is cosmically negligible
>evolutionary biases everywhere
>explain that moral intuition cannot be trusted
>except intuition that future digital people count
>and intuition that extinction is uniquely bad
>and intuition that our probability estimates are sane
>and intuition that our institutional choices improve the future
>those intuitions survived peer review
>someone donates $5k to local homeless shelter
>inefficient
>could have funded 0.0000000000003% of an AI governance researcher
>think of all the simulated people you just killed
>okay maybe don't phrase it that way publicly
>PR team says "future generations deserve a voice"
>much better
>journalist asks what longtermism means
>say "future people matter"
>everyone agrees
>great
>journalist asks what follows from that
>well technically we should redirect enormous resources toward low-probability interventions affecting astronomical futures
>journalist raises eyebrow
>return to "future people matter"
>motte has entered the chat
>critic: of course future people matter
>me: glad we agree
>critic: I don't agree that your institute knows how to help them
>me: why do you hate our grandchildren
>eventually notice uncomfortable implication
>if future value dominates everything
>then helping people today mostly matters through effects on future
>education matters because future institutions
>health matters because future productivity
>democracy matters because future trajectory
>human beings slowly become instrumental variables in their own moral philosophy
>see starving child
>feel compassion
>check spreadsheet
>child's direct welfare contribution negligible
>but perhaps childhood nutrition improves national institutional quality
>compassion restored
>tell myself this is impartial altruism
>one day assistant asks obvious question
>"how do you know your intervention actually improves the far future?"
>silence
>open spreadsheet
>increase column width
>add confidence interval
>assistant asks again
>"no, I mean how do you know the sign is positive?"
>stare into cosmic light cone
>10^50 people staring back
>none of them exist
>none of them can tell me
>none of them can falsify my assumptions
>realize I have invented the perfect constituency
>infinitely important
>completely silent
>and always represented by me
Sonovabitch, I'm in!
My explanation from reading their docs:
Extremely(!!) fast and accurate classifier - on the level of GPT 5.6 Terra - that takes a set of pre-defined options and a natural language query + data, and returns a solution from the options, with confidence percentages assigned.
In agent world, would be very useful for things like model routing, permissions classifier, safe / unsafe PR labeling, etc.
In production workloads, should be an extremely capable judge for any fuzzy logic that currently relies on LLMs, while providing stronger guarantees.
tldr: A switch statement on steroids, limited to about 255 options, that takes natural language queries and is as smart as Terra.
My explanation from reading their docs:
Extremely(!!) fast and accurate classifier - on the level of GPT 5.6 Terra - that takes a set of pre-defined options and a natural language query + data, and returns a solution from the options, with confidence percentages assigned.
In agent world, would be very useful for things like model routing, permissions classifier, safe / unsafe PR labeling, etc.
In production workloads, should be an extremely capable judge for any fuzzy logic that currently relies on LLMs, while providing stronger guarantees.
tldr: A switch statement on steroids, limited to about 255 options, that takes natural language queries and is as smart as Terra.
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
I can't help but once again point at art.
AI art has had the exact same glaring problem for years now. It produces the output without any of the intermediary steps that allow for better expression, artistic growth, and communicating something personal.
It's interesting that math is now hitting the exact same core problem from the other side. An opaque solution to a problem has value, but too much information that could be more useful is lost from the lack of intermediary steps.
A lot of people are missing Terence Tao’s point and thinking “mathematicians are upset that AI is better than them.” That’s not what he’s saying, and some people are forgetting that Tao is one of the most AI-pilled mathematicians out there.
His point is that when people work on discovering something, along the way they invent new concepts. Those concepts later become useful far beyond the original goal, and enables further inventions. Finding a solution does matter, but the intermediate idea is often what makes the field richer, because other people can share it and build the next thing from it.
In tech, we can use the analogy of collaborative software. We started with algorithms for merging changes in a Word document, and evolved that to concepts about versions, diffs, and merges, and later to real-time collaboration tools like Git, Google Docs, and Figma. Humans built upon these concepts and developed more powerful solutions.
Terence’s worry is that a machine automating a solution robs the field of the value of developing the intermediate discoveries in the pursuit of larger discoveries.
When automating a solution, the intermediate discoveries and invention of concepts can be buried or completely hidden in the black box. We don’t learn from them to build the next thing; it’s like we never made the invention of collaborative document editing and thus could not have the conceptual understanding to invent the next version – and since it’s hidden, we also don’t socialize them to allow other people to invent, too, a core tenet of collective discovery.
So then, in both code and math, this leads to the atrophy of development of concepts in the field.
In other words: pure ‘solution extraction’ that hides the process of discovery can leave the field with a checked-off theorem but little new insight or new questions to pursue. And it might prevent us from understanding a field deeper.
I am seeing, first-hand, that atrophying of skills in software development. We push buttons and get solutions. There is much less incentive to develop new concepts and human skill. The bet most software companies are making is that LLMs are so effective in writing code that you’re still shipping overwhelmingly more value even with human skill atrophy, and it’s the right bet IMO.
However, much of the software industry is built upon building things, not necessarily novel invention and research. In such an environment, you can say that you accept some atrophying of conceptual invention and human skill for more output.
On the other hand, sectors like math and pure sciences that are focused on invention and insight might be the hardest hit by this.
Practical/applied sciences might fall somewhere in the middle. An Alzheimer’s cure, room-temperature semiconductor, or highly effective carbon capture solution are far too valuable to sandbag and say only humans can do that to develop concepts in the ‘proper’ way. The outcome matters too much to treat the preservation of concept invention as the highest goal. Even there, though, hidden intermediates can slow the next breakthrough if nobody can see how the first one actually worked.
So the question is not “is AI allowed to solve hard problems?” It is “in this field (math, science, tech, etc.), is the answer itself the main point, or are the concepts and abstractions we use to get there also the thing we need to maintain?”
In pure math, there’s an argument that the intermediates are often more useful than the solution, and atrophy in concept development is highly detrimental to the field. Solving Navier–Stokes, contrary to what some people claim, has little practical application, and pure math might be one of those fields where just finding a solution isn’t the entire point, and can actually be contrary to the field, which is what Tao is worried about.
Ivan Tanev retweeted
📢 I'm officially looking for a new job. You hiring?
Full time, remote from Spain, paid via EOR, IC, TypeScript, no crypto, support for 3-4 conference speaking gigs/year.
Share plz. 🙏 DMs open.
Act now, while supplies last!
How is it that opencode can render tables and mermaid beautifully (and in general has the best rendering of all harnesses), while codex and claude can't even render text right :-/
Hey @mattpocockuk, how do I convince you to remove /tdd from /implement and replace it with smaller guidance around areas to test instead of TDD?
From my observation, /tdd degrades agent performance and generates a larger number of less useful tests (Fable, Sol, Astra)
I spent some time giving the same tasks to Astra and Fable 5.1
For coding Fable produces significantly better code; For simple refactors, Astra is much faster; My go-to stays Fable, though Astra has uses.
This might be a model training problem. Models are very happy to add code, but they're extremely bad at removing code or simplifying design. Some guidance helps, skills help, but the only true solution is:
Read the code carfully.
Tell the model the exact refactors to do.
You need The Comment Sicko: github.com/cursor/plugins/bl…
Very easy to pull into an isolated skill, highly recommend.
