> My bet is their model doesn't realistically compare to any of the frontier models.
I've been using composer-1 in Cursor for a few weeks and also switching back and forth between it, Gemini Flash 3, Claude Opus 4.5, Claude Sonnet 4.5 and GPT 5.2.
And you're right it's not comparable. It's about the same quality of code output of the aforementioned models but about 4x as fast. Which enables a qualitatively different workflow for me where instead of me spending a bunch of time waiting on the model, the model is waiting on me to catch up with its outputs. After using composer-1, it feels painful to switch back to other models.
I work in a larg(ish) enterprise codebase. I spend a lot of time asking it questions about the codebase and then making small incremental changes. So it works very well for my particular workflow.
Other people use CLI and remote agents and that sort of thing and that's not really my workflow so other models might work better for other people.
Does it have some huge context window? Or is it really good at grep?
The Copilot version of this is just fucking terrible at suggesting anything remotely useful about our codebase.
I've had reasonable success just sticking single giant functions into context and asking Sonnet 4.5 targeted questions (is anything in this function modifying X, does this function appear to be doing Y) as a shortcut for reading through the whole thing or scattershot text search.
When I try to give it a whole file I actually hit single-query token limits.
But that's very "opt-in" on my part, and different from how I understand Cursor to work.
It is really good at grep and will make multiple grep calls in parallel.
And when I open it in the parent directory of a bunch of repos in our codebase, it can very quickly trace data flow through a bunch of different services. It will tell me all the files the data goes through.
It's context window is "only" 200k tokens. When it gets near 200k, it compresses the conversation and starts a new conversation..... which mostly works but sometimes it has a bit of amnesia if you have a really long running conversation on something.
When other models would grep, then read results, then use search, then read results, then read 100 lines from a file, then read results, Composer 1 is trained to grep AND search AND read in one round trip
It may read 15 files, and then make small edits in all 15 files at once
Presumably if it knows it needs to perform multiple searches in order to gather information (e.g. searching for redundant implementations of an algorithm, plus calls to the codebase's canonical implementation) it should be able to run those searches in parallel grep calls.
LLMs are inherently single-threaded in how they ingest and produce info. So, as far as I can gather from the description, either it spawns sub-agents, or it has a tool dedicated for the job.
In my experience the software engineers that often get promoted crank out as many features as possible with almost no regard to the defect rate. If you do things at a more deliberate pace, you can be regarded by product or management as slow and unproductive. Real incentives exist in software development that push for quantity over quality.
This is not analogous to the failure of an anesthesiologist to properly sedate a patient. This is a process failure. Clearly the proper amount of QA was not in place for whatever reason and they need to re-examine their approach. Of course their process needs extra diligence given the cost of failure.
It would be a real shame if an individual is punished for this instead of examining the process and system of incentives that led to this failure.
>Walmart will pay for workers in those locations to transfer to other primary offices, such as San Bruno, Calif., or the company’s headquarters in Bentonville, Ark. The company hopes to relocate most of the workers, and some will be allowed to become full-time remote workers, a spokeswoman said. Those who leave will be given severance pay, she said.
It looks like they had 17 techhubs worldwide, and are closing 3 in the US. The remaining US ones are Atlanta GA, Bentonville AR, Charlotte SC, Dallas TX, Hoboken NJ, Reston VA, San Bruno CA and Seattle WA.
Given they just opened new locations in Atlanta and Seattle last year, this sounds more like Austin and Portland are getting too expensive (and aren't as essential of a labor market as SV), so they can save money by focusing on other locations.
I assume you're being droll but to me (not a Walmart person) it seems like Bentonville would be better professionally since it's company HQ, and very much a company town? My completely unfounded theory is that it's generally a more savvy career move for your boss to see your face IRL? The thought is that if two candidates are equal, the one that actually shows up in the office is just going to be more likely to get the nod.
You've never been to Arkansas... There is much more to factor in. If you have a family, it's literally one of the worst places to live/work. The median pay is under 30K, and it's almost dead last in health care and education. There are about 45 other states where you would be better off.
LITTLE ROCK, Ark. – Based on a recent report, Arkansas ranked next to the bottom on good states for millennials to live in.
In a recent report by Scholaroo, all 50 states were ranked from best to worst states to live in as a millennial.
Study ranks Arkansas school districts on equity, ranks state #2 in US
Each state was judged on multiple metrics, from affordability and safety to their political and social environments.
With two exceptions, Arkansas ranked near or at the bottom of all the metrics with an overall ranking of 47.
At its highest, Arkansas ranked 13th on affordability of living and 25th in employment. Otherwise, it was mostly rock bottom.
Arkansas ranked 42nd on its political/social environment and 49th in quality of life.
At the absolute worst, Arkansas ranked 50th health-wise.
===================================
I would have to be offered a truly absurd amount of compensation (as a W2) to live and work in a Walmart office in Bentonville, AR.
I don't have kids but if I did there is no way in hell that I would want them attending any AR public school system.
My lumber dealer gets his cedar cants from Arkansas, so that would probably cut out more than half of my lumber costs and save me around $600 per year currently, but if my CoL dropped so significantly, I could probably retire to furniture making a lot sooner too.
I'd have them calculate the post income tax cost of a daily plane ticket to and from Arkansas for my commute, plus the cost of all other transportation and meals, add that to 150% of my current salary, and then we can talk about whether the job opportunity makes sense.
Walmart and the Waltons dump gobs of money on Northwest Arkansas. It doesn’t make much sense to compare it to the rest of the state. Bentonville is a pretty nice place to raise a family. Public museums, schools, parks, bike trails, all really nice.
If you plan on being a Walmart lifer, sure. But what about your next job if that isn’t true?
Better to be central to where the majority of companies and candidates are, where your ex-colleagues, parents of your children’s friends, and casual acquaintances (proverbial bowling league) all are part of a broader and accessible network for future job prospects.
It also keeps you culturally (professionally) more similar to your next job’s culture.
I know someone who had a non-technical corporate management role in Bentonville. The company basically decided to eliminate the group and he needed to move somewhere else to even job hunt because there sure wasn't anything else locally.
Setting aside the dubious merits of Arkansas (and yes I've been to Bentonville), why would a technology person move to where he'll never rise up that high in the greater organization? I think the answer to your question, is to recognize that you shouldn't put yourself into such a position.
Because internal promotions are so rare these days in tech. You have a much better chance at getting a promotion by getting another offer from a different company.
Prices are already highly speculative in that region and this move by Walmart will further fuel speculators who think will try to squeeze a quick buck flipping or renting to a new wave of captive out of town techies.
Yeah I was a tubist. Timpanis were arms length away, I have to assume that the effect is only for French Horn players…that tiny mouthpiece and the continuous tapering and the intense flare at the end…it’s also aimed a funny direction compared to the other brass instruments (a little to the side rather than up or straight ahead).
BMC Group K.K. Japan | Front-end Web Developer | Remote Asia and Oceania | Contract | $ Hourly | Flexible schedule
We're a small distributed team of 3 core developers, 2 support service configuration specialists / BAs, and a PM, looking for a 3rd core developer to help us build an insurance application, and improve our existing technical products and services.
Strong communication skills in English is a requirement.
Knowledge of Japanese is NOT in any way a requirement.
You must live within 3.5 timezones of Japan (JST - GMT +9). Countries within these constraints include India, China, Sri Lanka, Bangladesh, Thailand, Myanmar, Cambodia, Laos, Indonesia, Singapore, Malaysia, Brunei, Vietnam, Philippines, Australia, New Zealand, Japan, Taiwan, South Korea, parts of Russia, Papua New Guinea
What you might do:
* Write code to create robust web application front-ends utilizing React, Redux, TypeScript, and HTML/CSS.
* Translate user stories into wireframes and then turn them into mockups before implementing them as code.
* Continuously optimize web designs to provide a smooth, beautiful user experience.
* Troubleshoot bugs and issues that may require investigating both the front-end and back-end.
* Closely collaborate with back-end and full stack developers.
Required experience/technologies: ES6+ features (Promises, Async/Await),
Web UI design with HTML/CSS, Front-end JS libraries/frameworks (React, Angular, Vue, etc.), Single-page applications (SPA)
Preferred experience/technologies: Typescript, React/Redux, Docker-based development, General understanding of back-end web frameworks, familiarity with build tools like Gulp, Webpack, and Parcel
Bonus experience/technologies:
* Node.js
* Enough back-end experience to implement simple endpoints with queries
* AWS or other cloud services (for deploying web-applications to development environments)
Please contact evan.scholl@bmc-group.co.jp with your resume. Not required, but side projects and github contributions are helpful and appreciated .
Complete extinction of humans is a high bar. Also I would expect that the reaction to such a disease would be extreme. It could certainly be much worse than the current crisis we are facing.
You are assuming fund managers want to maximize returns for investors. That is not the game they are playing at all.
Fund managers have pathological incentives. People won't hand them money if they say 'I'm going to just hold this money as cash for the next 1-5 years until the market goes completely irrational in a way that I can make you 5x returns, but in the meantime I'm going to charge you 2% per year on the balance'.
They have to put that money in to something or they won't be given money in the first place.
Secondly, funds play games with the money, because investors are not sophisticated and there is a ton of dumb money in retirement accounts. People look at the past returns and choose funds that have a high past rate of return. Fund managers know this, so they set up 15 funds with different strategies. One of them has a high rate of return, by luck, while 14 do much worse. Then they expand that fund 10x by quoting that high rate of return.
Fund managers can make more money by not optimizing how much money they make for you, if you measure them by how much money they get paid to manage other people's money you'll see they are playing that game very well.
Note that report is Vanguard marketing material and most managed funds have specific mandates (like maintaining a certain volatility or investing in certain securities) other than maximizing gains.
This is why hedge funds underperform indexes in bull markets but beat them in turbulent times.
> This is why hedge funds underperform indexes in bull markets but beat them in turbulent times.
Agreed on the first part; I’d want to see hard statistics (after fees) on the second part (and, no, citing the 3 or so famous exceptions that are closed to outside investors and might have used illegally obtained insider information doesn’t invalidate the larger point).
Hedge funds trade gains for consistency. They don't publicly report numbers though and aren't meant to be public funds (that's what ETFs are for) so it's hard to get full stats, but it's common knowledge in finance.
That aligns with what I said, although there are plenty of private hedge funds that don't report their returns. The benefit varies for each client.
An example would be a real-estate company that has holdings in a fund. Preservation and low volatility is far more important than raw gains and can be used as collateral to offset losses in physical property valuations.
>Very few active fund managers can consistently beat their indexes.
Funds profit from managing more money. The more money you manage, the harder it is to find plays that use a significant amount of it. So it makes perfect sense that funds would increase in size until they perform at a rate similar to indexes.
>Why do you believe you would be better over a long period of time?
The fully loaded cost of even a junior analyst would be in the hundreds of thousands a year. Independent day traders can focus on strategies that leverage such a tiny amount of capital and result in such a tiny return that it's not profitable for a fund to be looking at. Keeping in mind that "too small for a fund to worry about" can still be "shitloads of money" for 99% of people.
Beating indices in a correlated bull market isn't the same thing as making money consistently.
The contrapositive of your argument is that anyone who is holding cash outperformed the S&P today. Though true, it sounds silly to say that someone who goes about his day with a $100 in his wallet is outperforming the market, no?
In a disperse market, single stocks chosen correctly will outperform the index. In a correlated bull market, the index will outperform.
I would agree with you that there hasn't been a genuine hit since MOO 2. I would say, however, that one game, Distant Worlds: Universe, was a huge leap forward in redefining what's possible in the 4x genre (i.e. deep and immersive gameplay in a real-time context). Most of the other 4x games since 1996 have been a variation on the gameplay mechanics of MOO 2.
DW:U is the only 4x game that I would unequivocally say is as good as the original MOO2.
I've been using composer-1 in Cursor for a few weeks and also switching back and forth between it, Gemini Flash 3, Claude Opus 4.5, Claude Sonnet 4.5 and GPT 5.2.
And you're right it's not comparable. It's about the same quality of code output of the aforementioned models but about 4x as fast. Which enables a qualitatively different workflow for me where instead of me spending a bunch of time waiting on the model, the model is waiting on me to catch up with its outputs. After using composer-1, it feels painful to switch back to other models.
I work in a larg(ish) enterprise codebase. I spend a lot of time asking it questions about the codebase and then making small incremental changes. So it works very well for my particular workflow.
Other people use CLI and remote agents and that sort of thing and that's not really my workflow so other models might work better for other people.
reply