Chyzwar

Are you suggesting that European should use third rate models and accept competitive disadvantage?

I completed my master while working full time but. It was though but necessary as my undergraduate degree was not enough to land software job and I lacked real skill.

I would probably take a job. If AI bubble pops you would be debt free and with 1-2 years of experience.

Ask Claude cowork to click for you.

I think the best way to test this would be to put PgBouncer or a similar proxy in front of a busy production database, and mirror queries to both traditional Postgres and the Rust one at the same time. Then you can compare output and performance under real load. After running it for a while, you could diff the tables one to one against the normal Postgres instance.

"If I had asked people what they wanted, they would have said faster horses." Henry Ford.

He diluted his vision to garbage on the first enterprise customer. Because he needed to sell now. From OP, there is visible urgency to do sales even when the product is not ready.

Tesla was overhyped in the first decade after acquisition by Musk; it was bleeding money for 15 years. Arguably, the first Tesla cars were worse and pricier than traditional cars.

My point is that compared to the US, in Europe you have very few stories of successful stealth startups or startups that have needed lots of money for many years. That is a main reason why we cannot create a frontier LLM lab.

This is such European take on startups. Tesla was making shitty overpriced status symbols/value signalling cars and selling FSD for 10k knowing very well that it will not work with car hardware. It took them 10 years to "fake it until you make it stage".

If founder keep iterating and hyping his ovens with enough capital he could become big player in oven maker space and disrupting industry. Learning from this article was that he lacked capital and vision.

My 7-day stats from the codeburn tool: $2,123.26 cost, 11,370 calls, 333 sessions, 96.7% cache hit. Breakdown: 4.5M in, 18.1M out, 2,084.9M cached, 66.6M written.

I spent $2,000 in tokens on my $200/month subscription in just one week. This also excludes heavy usage of Cowork and Claude Design.

Claude estimated that the same usage on DeepSeek Pro API would cost me $200 for the week. A month averages 4 weeks, so I can estimate that Claude is 4x cheaper than DeepSeek for me.

What are you basic your opinion ?

They should better manage expectations. They put rootless as a forefront before it was ready and still no proper support for Ubuntu on v4 of Podman.

[1] https://github.com/podman-container-tools/podman/discussions...

Last time I evaluated podman, Ubuntu was second class citizen. Rootless was non trivial and required additional setup. Documentation also suck.

Docker is something we all already hate, milion edge cases and forever bugs but at least well documented and understood. Podman claim to be drop-in replacement does it mean it carry docker shitness? Examples: ufw punch through, env file handling, volumes, etc

Until 7th,Fable is twice expensive in subscription tokens than Opus. They are testing if they can introduce 400 dollars Fable subscription.

npx codeburn

Because claude subscription tokens are cheaper than deepseek and friends. You have whole industry of people reselling Claude subscriptions in China.

Can you comprehend than Anthropic is winning because is both cheap(subscriptions) and better SOTA. People are cheering China providers when I reality they would rugpull open weights the moment they are competive.

China models are trash that why they are giving them away for free.

For individuals and small companies subscriptions is the best deal, for big companies china models are big no unless they can host them.

1. But it might last for at least few more years, see Nvidia 1 trillion backlog.

2. Semiconductor manufacturing is the most complex industrial process in the world. You need billions of capex and decades of experience. Even existing semi players like intel cannot switch production to memory.

China CXMT is gaing traction in DDR market. New fabs from all players wil come online in the next two years.

AI makes result oriented developers with enough agency to absolutely crush metrics and work. On the other hand, detail oriented, ticket spoon feed developers look pathetic for not being able to make same impact. Previously these two groups could co-exist were detail oriented developer would keep codebase stable and result oriented would spike new features. Now you can just spin an agent.

What they wanted is to regulate competition but they forgot that current admin is Trump. They should have buy gifts to Trump preferably gold and pay for few exciting dinners in Florida. They should have learned from Bezos and Cook.

Enterprise use Teams, mattermost, nextcloud talk, Google meet and many others.

No, Teams is just better product. It handle large video meetings, has build in recording and calendar integration. Authentication is seamless and handling many workspaces is better. Slack is only better in chat interface.

Why Slack? Everything about this product sucks: bad video calls, no meeting recording, limited screen sharing, janky authentication, and predatory pricing.

They shoudl just expose SDK/platform for people to build own integration with Discord/Teams/mattermost etc. This would allow for fine grained permisions control and speed up adoption. To large degree this is alternative for OpenClaw ?

You can only take responsibility if you also have ownership. Large organizations very quickly become like communist state, nobody own anything and any individual action is suspicious.

https://arxiv.org/abs/2604.24827

From this paper

When discussing LLM pricing, people are missing the plot. The subscription token price is 10x-40x cheaper than API pricing. Your 90$ Claude subscriptions give you close to $1000 to $4000 in equivalent API token pricing.

The second issue is that the quality of the model “operator” makes a massive difference in the outcomes. Highly skilled senior devs who know how to prompt and have high agency will outperform team people that lack motivation and foundational skills.

Lastly, there is a massive difference in capabilities, determinism, and error handling between 5T SOTA models like Opus and tiny distillations from DeepSeek that perform well only in benchmarks.

You can fix Claude's laziness by modifying the system prompt. https://gist.github.com/chyzwar/99fe217c3ed336f57c74dcffe371...

Gross margin calculated on API token pricing with discounted training and hardware deprecation.

I am no so sure about batch sizes, chatgpt napkin calculation for 5T model show 10-300 sessions.

All SOTA model providers are losing money. When users run Opus, they are essentially renting a GPU cluster worth half a million dollars for a $100/$200 subscription. If they want to IPO, they need to show a projection toward profit. For that reason, they want to discourage power users and attract normies.

Because your fetch most likely mishandle errors, lack retries, fail on redirects and is unnecessary verbose for both people and agents.

At lot of advice for solo founders is from people with survivalship bias, vocal sales biz dev and people that had right connections. I think tech solo founder path is different and in today market is very difficult. I can give advice on things that I did wrong so far. Avoid direct stripe integration and complex subscriptions models this would be time sink, use merchant of record and as simple as possible subscription model. Ignore everything non essential for product like usage charts, account management etc. Get decent designer but limit his scope to product features say NO to design system and custom components, use off the shelf UI toolkit with minimal customisation. If you want to raise wait for more traction and avoid startup conferences, better to go for your industry conference instead.

Exactly without billions in investments this would yet another experimental ISA.

I recently tested Claude code, opencode, codex on same frontend feature and codex with 5.4 with high effort was the best but most expensive. For me in Europe, Claude Code with 90$ max subscription is the best value for money.

My thinking is:

  codex - best harness
  opencode - best ux/dx
  claude - best value for money

This could be controlled by npm. Client ask for available versions anyway. If package is security fix then it can be made available instantly. But this delay gives time for security scanners and time to notify maintainers that package was published.