iandanforth

This is search. Please don't reinvent search with a new acronym.

Agents can use Google. They can also see all the same signals humans do to judge quality. Maybe there's a need for a specialized directory, but does it need a spec or an acronym? No.

Not a physician. Some observations on these statements.

The predicate is "given how we practice medicine and the limits of humans ability to interpret the imaging modalities we have."

The more specific predicate is "for my specialty would this replace or prove superior to the tools that I have?"

Both of these are totally reasonable, however the history of medicine, and science in general, is that creating new ways to look at things has a tendency to reveal information that we never knew we needed.

For example, for years I thought of blood sugar as something that was either in a good or bad range. Then I tried a continuous blood sugar monitor. The full picture of the body's response to specific foods that I ate was eye opening. There's so much more to learn when you get a higher resolution (temporal in that case) view into your body.

Another wonderfully hopeful example is the retinal imaging ML work done by google. A completely non-invasive image of the retina for diabetic issues, that also happened to be able to predict things like age, sex, smoking status, previous cardiac events and more! Just take high-res pictures of things! The body is interconnected in ways that you can infer from one system so much about others.

So while I don't think anything the Dr. said is "wrong", I think it represents a very common blinkered mindset of pragmatic practitioners who need to deliver reliable performance daily.

Funny you mention that, since that's what I pilot LLMs to build :)

Shutting down the growth prospects of a company based, not on its behavior, but on the capability of its models right before the IPOs of the companies you're going to profit from is staggeringly dumb. Yes the public is stupid when it comes to investing in stocks, but come on. If these companies growth prospects rest in large part on continuing to improve their products and the government said that if they do they face National Security Cease and Desist letters, then investing is a bad idea.

The selfish / corrupt thing to do is to do this after you've fleeced the public.

"We received the directive from the government today at 5:21pm (ET)"

This sounds exactly like the opening line from an apocalyptic sci-fi film.

Brutal. I love it.

Ah, ok. There's a whole body of literature here that I think divides our opinions. I would recommend "Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict (Columbia Studies in Terrorism and Irregular Warfare)" to start. The history of effective change in the face of organizations acting in bad faith may not be what you think it is.

Sorry if I was unclear. I don't work in finance. I do work with agents. I think expert engineers in finance who are guiding agents are adding a lot of value because of their knowledge of finance. Because I lack that knowledge of finance, even given access to agents, I would not accept a role guiding agents in a finance company because I wouldn't be able to guide the agents well and my/our output would be bad.

"I ended up working in software development roles in the domains of finance, bookkeeping and payment processing, where I had great autonomy and a close and candid relationship with Product Managers and stakeholders.

I learnt a lot about the domain and how to effectively write programs for it: PCI compliance, double-entry ledgers, escrows, reconciliation, payment lifecycles, bank transfer idempotency, etc.

It was, then, obvious that I should focus my career on becoming an expert on that domain to stand out as a professional and differentiate myself in a field that showed signs of an increasing need for domain specialists."

The point being that we're beyond where that's a responsible choice. Empathy for an organization enforcing its rules above the actions of those protesting them means either an ideological alignment with the censorship or an ignorance/disbelief of the severity of the harm the organization is causing. The former audience will never be persuaded. The latter require education and persuasion, and while its useful to create a sense of martyrdom via forcing the enemy to act in an obviously unreasonable fashion, it's a waste of time to argue with their definitions of rules. If the rule were "you are not allowed to say the governing body is corrupt" and they say its corrupt, that exposes clearly and plainly the problem, and enforcement of the rule provides no authority because the rule itself is obviously designed to quash dissent. If the listener is so blithely oblivious to how the intent of a rule has been manipulated to quash dissent, as it has here, then there is no loss in squarely addressing that. "We are protesting your abandonment of scientific principles" is both what they were doing and should be doing.

Wut? I pilot LLMs all day but there's no way in hell I'd agree to be at the helm of a finance product. That first pillar is still there. Maybe the author isn't aware of the impact they have, but I know, with the evidence of reverted PRs, that when I step outside my area of deep knowledge I can no longer call BS on the agents. Our most capable agent, with access to the same kind of distributed systems the author talks about, is regularly wrong, frequently myopic, and just outright dumb constantly. It's the expertise of engineers on the team that push it back on track.

A reminder to anyone who finds themselves in this kind of situation, do not engage with the rhetoric of the enemy. You cannot win an argument where they set the rules. So here, where they question whether or not they were "protesting" distracts from the reality of a censorious organization that will weaponize regulations it controls without good faith. Instead you need a simple, memorable statement of condemnation which is repeated consistently and a clear action which those who hear it can take in response.

"This organization is controlled by Trump loyalists. They are not scientists. You do not owe them respect. Speak over them. Let no manipulation go unchallenged or derided."

Claude can directly drive Codex or Codex can drive Claude. Both already produce logs. It's unclear what value this intermediary brings.

This is a great move for OpenAI and one that should worry Anthropic. Bedrock was the only way I could use foundation models for a while given AWS lock-in and security requirements.

Putting two adaptive dynamic systems next to each other is tricky. Your eyes and these glasses could easily create a positive or negative feedback loop or begin oscillating. So while cool I hope they have some experienced controls people on staff to detect and prevent such things.

I have similar projects in mind. How were these printed?

Deepmind hype is the worst hype. They do genuinely cool stuff and talk about it publicly, but don't make it available. Or it's only available to a tiny select few. Just shut up about the things you're doing until they are ready. You're part of a consumer products company, not a university PR department.

Huge fan of zoning reform, however I'd love to see equal effort in lending reform. The availability of multi-million dollar mortgages on 30 year terms means that we all get poorer. Getting people into owned homes is a dream left over from the Clinton era that has warped into an ever expanding pool of debt and over sized buildings. Developers will build to the limit of what people can afford (and slightly beyond), and that is defined by mortgage policy. The harsh reality is that as long as you're supporting a system where a 1k sqft house can cost 300, 400, 500k you're not helping anyone who isn't in the business of lending. The only way to reverse the trend is to limit the pool of available capital and bring the sale value of property down.

I don't understand how people can use the phrase "right-size" without a crushing sense of embarrassment and shame. Did you swallow a business consultant from 1990? That and phrases like "go forward strategy" say either 1. I do not know how to communicate like a human or 2. I am afraid of speaking naturally because it impinges on my self image as a business leader or 3. I do not want to accurately describe what I'm doing because that might expose my fragile ego to the possibility that I'm doing something which hurts people.

"We're firing a bunch of people because we think we don't need them anymore due to AI and we'll make more money without them."

There are times when businesses must fire people to stay afloat and it's a business that objectively needs to exist. This isn't one of them, so don't waste everyone's time with your BS, please.

I found the section on publishing very interesting. Even if the quality of the output is up to snuff, where should it go? Arxiv doesn't allow AI written work. The author proposes that only work that has been certified by human should be published. However, now the field is in the same boat as software engineering where we are facing a glut of pull requests and not enough time and people to review them.

Try this thought experiment. If, in 6 months, the agents were better coders than you are, would this argument still hold?

This is a personal thought experiment so think it through for yourself. What would the consequence be if the agents really were better than you and you acknowledged that?

The major premise of "It's a trap!" is that it matters if you lose your coding skill. (I'll gloss over general critical thinking and stick with coding for now) However in the world where on any given task it would be done to a higher level of quality and faster if you gave it to the agent, then what are you doing trying to do it yourself? There's plenty of room for that kind of thinking in hobbies, but in the professional world?

Maybe you can add some value in code reviews, but you may also be better off never reading the code at all. Maybe the how of coding stops mattering and the what of products needs to be your top concern.

I can tell you that the agents that I use today are much better coders than I am in the language we're using. I don't write it at all. I couldn't fizzbuzz in it. But with a small team we are building useful internal tools and features at a breakneck pace. I certainly feel the same feelings of getting dumber and losing my coding chops, but I have to step back and say, could what we've built have been built in 5x the time without agents? And the answer is probably no.

The thing I'm mastering now is conjuring software with agents. What lets them rip, what slows them down, where they are today and where they will likely be tomorrow.

I can tell you that you should re-invest in small, modular systems, because agents can build modules and greenfield projects instantly. I can tell you that there is a point at which agents fall over completely even on mid-sized projects, but that that point is receding with each new generation of model, and that Codex 5.4 XHigh Fast set to 500K context window is a beast. (5.5 has yet to win me over)

I can tell you that pushing direct to main is viable, that PRs slow down fully agentic teams, and if your agents have sufficient permissions they can fix things fast enough to be let loose even knowing they may delete your service. I wouldn't do it with your main product yet (unless you're starting your startup today) and I wouldn't try it with a large legacy project. But maybe that rewrite you've always wanted to do is here and just a prompt away.

Now, the sane among you will note that agents are not better today, that they might not ever be, and either way you should never trust a computer to make a decision because it can't suffer the consequences of its actions. Or more down to earth, there are some things that are too important to yolo.

But I will argue that a huge swath of us work in domains where if you're willing to challenge some of the basic assumptions of software development (you should understand the code, it should be maintainable by humans, it should be built to last) then you'll be able to provide very useful software much more quickly than you would otherwise be able to do. Save the skill for your hobbies, and build things people want.

Anyone know if these are already powering all of Gemini services, some of them, or none yet? It's hard to tell if this will result in improvements in speed, lower costs, etc, or if those will be invisible, or have already happened.

First I'm hearing of them and with this ownership I'll be highly skeptical of any of their content if I do happen to watch.

I like this idea! I don't need the LLM bits, and want it to run on an old Android tablet I have lying around. Can anyone recommend similar software where I can get wikipedia / street maps / useful tutorial videos nicely packaged for offline use?

I'm very happy about this change. For long sessions with Claude it was always like a punch to the gut when a compaction came along. Codex/GPT-5.4 is better with compactions so I switched to that to avoid the pain of the model suddenly forgetting key aspects of the work and making the same dumb errors all over again. I'm excited to return to Claude as my daily driver!

I don't think we won't get AGI if Anthropic were to implode, and frankly, right now, I'd rather have someone say clearly, "They cannot stomach the existence of someone telling them 'No' or adhering to moral principles. Like spoiled children they can't hear the former and are terrified by later because it might expose them to the condemnation they deserve."

TLDR (story, not math) - Knuth poses a problem, his friend uses Claude to conduct 30 some explorations, with careful human guidance, and Claude eventually writes a Python program that can find a solution for all odd values. Knuth then writes a proof of the approach and is very pleased by Claude's contribution. Even values remain an open question (Claude couldn't make much progress on them)

Totally reasonable project for many reasons but fast tools for AI always makes me chuckle. Imagine your job is delivering packages and along the delivery route one of your coworkers is a literal glacier. It doesn't really matter how fast you walk, run, bike, or drive. If part of your delivery chain tops out at 30 meters per day you're going to have a slow delivery service. The ratio between the speed of code execution and AI "thinking" is worse than this analogy.

The crucial thing is that Tesla's valuation has the hype projects baked in. The fact that it never delivered self driving or a robotaxi fleet and is now being saved solely by an import ban on Chinese EVs means that any success he had with Tesla is now an illusion.

While this seems to detect posture fairly well, the screen blurring doesn't work for me despite allowing what appear to be the relevant permissions. (macOS 15.1)