personally,
i did one test of ling3-flash and gemma-4-31B side-by-side .
ling3 understood me and had a fantastic answer. gemma must have misunderstood what i was saying… it wrote a long story-like paragraph that didnt answer my question. but it kinda had 1 bit of insight.
so: (maybe…) don’t sleep on ling models !

i think a key consideration is wether there are alot of developer for the lang.
a foss project will grow faster if its written in rust,js,cpp,python

my last 5 uploads averaged 10kilobyte per img (webp).
i don’t disagree to say that we shouldn’t use the best technology available.
but relevant to note: proper optimization of the image goes a loooong way.(often more effective than just relying on a file format with theoretical-max efficient compression.
i love this subject… because i hate reddit where ppl upload a screenshot of plain-text from twitter and it costs me 1MB of scarce mobile data, to see some garbage post. fuck that

hurricane harvey was pretty bad. 50k-100k crisis refugees .

thank you. great post
https://vger.to/piefed.social/c/best_fdroid_apps/p/2167772/fastlog-for-editing-your-logseq-notes-on-older-phones

really a gamechanger for logseq

“the internet” is a super broad term.
i had cancelled my home ISP to save cash. really a smart move IMO.
i think everyone(or technically, most people)… need some basic internet for some specific uses.
a mobile data plan should be the only thing you really need.
you could (ideally) talk to your neighbors and negotiate to share their wifi.

chinese phones seem to have the highest available ram , for some reason…

also should appreciate seeders who are the sole seeder for (otherwise) dead/lost torrent

AI critique(criticism… focusing on missing angles)

Like: The Button That Changed the World is strongest as an insider-led history of how a tiny interface feature became a business, cultural, and behavioral force. But its scope appears more explanatory and innovation-focused than a sustained critical investigation of the harms and power structures the button helped enable. manoflabook

What it seems to underexplore

• The Facebook “Like” widget beyond Facebook. A reviewer specifically faults the book for not adequately addressing Facebook’s external-web Like button, which extended tracking and helped sharpen Facebook’s advertising and algorithmic capabilities beyond activity inside its own platform. manoflabook

• Surveillance and data extraction. The book discusses profiling, marketing, and the value of likes as data, but it seems less focused on the political economy of surveillance: consent, opaque cross-site tracking, data brokers, and the asymmetry between users generating behavioral data and platforms monetizing it. netgalley

• Mental-health effects with scientific depth. It draws on psychology and neuroscience, but a critical review could ask for more rigorous treatment of social comparison, compulsive checking, adolescent well-being, body-image pressures, and the distinction between correlation and causation in research on social media harms. The broader discussion around the button explicitly connects its feedback loops to mental-health concerns, but that issue does not appear to be the book’s central analytic project. store.hbr

• Misinformation and political polarization. Likes give ranking systems an easy signal of attention and approval, which can reward emotional, divisive, or sensational material. The book’s business-and-innovation framing may leave insufficient room for detailed cases involving election influence, coordinated manipulation, extremist content, and the design choices that convert engagement into reach. imd

• “Like” as a measurement problem. A like can mean “I agree,” “I saw this,” “I support you,” “this is funny,” “I am being polite,” or even “this is terrible but important.” The book likely recognizes the button’s ambiguity, but it could go farther in examining the consequences of platforms treating this context-poor gesture as a reliable preference signal. youtube

• Who bears the costs. The book seems centered on creators, companies, platforms, and innovation. A fuller account would foreground groups with less power: children, gig workers rated by customers, marginalized communities facing coordinated harassment, journalists whose incentives shift toward engagement, and small businesses dependent on opaque algorithmic distribution. store.hbr

• Alternatives and reforms. It apparently tells why the button succeeded, but may devote less attention to what should replace or constrain it: chronological feeds, friction before sharing, private feedback, non-public metrics, age-sensitive design, interoperability, data minimization, and regulation of targeted advertising.

A useful critical thesis

You could frame your review this way:

Reeves and Goodson convincingly explain why the Like button spread: it reduced the effort required to express recognition and generated an extraordinarily valuable data signal. Yet the book is less complete as an account of the button’s social costs, because it gives comparatively limited attention to how that signal became infrastructure for surveillance, algorithmic amplification, political manipulation, and status-driven behavior.

That critique is fair precisely because the authors present the button as having a “profound impact on modern human interaction,” not merely as a clever product innovation. store.hbr

https://vger.to/lemmy.ml/post/50758732

create a sub. write a sidebar with definition, resources, and guidelines. then moderate it for the greater good

lemmy is no different in terms of AI training well… a little better. but not a blanket of safety

low-level compilers can output very ugly-looking assembly. he probably did this and then used LLM to super-optimize it. may be perfomant, but id guess that theres a risk that its unsafe.

it would help investigators if you get murdered

on the extreme opposite side of the spectrum:
https://huggingface.co/LiquidAI/LFM2.5-2.6B
or… really extreme… https://huggingface.co/AxiomicLabs/GPT-X2.5-135M

personally, i would be cautious with discord too

ai summary

Summary

The United States’ former focus on “can we stay ahead of China in AI?” has been replaced by a new reality: China is no longer just catching up, it is building an entire AI ecosystem that competes with the U.S. across performance, cost, deployment, financing, standards, developer adoption and global reach.

Key points

• China’s AI surge is ecosystem‑wide. Companies such as DeepSeek, Moonshot AI, Alibaba, Tencent, Zhipu AI and MiniMax are not isolated successes; together they show a coordinated, repeatable ability to produce world‑class models.

• Washington’s response is lagging. U.S. policymakers continue to treat each Chinese breakthrough as a discrete event, while China pursues a long‑term, systematic “ecosystem statecraft” strategy that integrates industrial policy, finance, standards, education, diplomacy and commercial expansion.

• Ecosystem statecraft vs. company‑by‑company competition. The U.S. still relies on frontier innovation and export controls, but China is reshaping the whole technology stack—making AI easier to deploy, customize and integrate, and encouraging worldwide developer adoption.

• Strategic intent. President Xi’s calls for AI cooperation, open‑source development and involvement of developing nations signal Beijing’s aim to become the architect of a global AI ecosystem, protecting core capabilities at home while exporting its stack abroad.

• Policy implications for the U.S.

  • The U.S. must move from a company‑centric debate to a national strategy that builds a competing ecosystem—combining research, standards‑setting, talent pipelines, financing, trusted alliances and diplomatic credibility.
  • America still holds major strengths: world‑class universities, a vibrant venture‑capital market, a dominant semiconductor industry and frontier research labs. Yet, historical precedent shows that lasting leadership depends more on who creates the adoptable ecosystem than who invents the first model.

• Global adoption dynamics. Nations are now weighing security, cost, financing and long‑term reliability rather than merely choosing between U.S. and Chinese hardware. Trust, developer communities and standards have become decisive competitive advantages.

• Conclusion. The decisive question for the United States is not whether its firms can keep building the most capable models, but whether it can marshal a coherent, resilient national strategy that yields an AI ecosystem that the world chooses to trust and build upon.

i distilled this article

Summary of the article “How China gets better bang for its buck than America in AI” (Aug 3 2026)

• U.S. AI spending is massive – Bloomberg Intelligence estimates U.S. data‑centre capital outlays could exceed $740 billion in 2026, with Nvidia alone negotiating a $250 billion financing deal for a $500 billion data‑centre run by OpenAI. Alphabet announced a $205 billion AI budget.

• China spends far less – Chinese tech firms are projected to invest less than one‑tenth of the U.S. amount in data centres. Yet their models perform only slightly behind U.S. equivalents. For example:

  • K3 (Moonshot AI) scores ≈ 95 % of Anthropic’s Fable 5 on common benchmarks while being 70 % cheaper to run.
  • Alibaba’s newly released model ranks among the world’s best on certain metrics.

• Why Chinese spending is efficient

  1. Lower input costs – Land, construction, equipment and labour are cheaper in China.
  2. Model distillation – Chinese labs often train models using outputs from expensive U.S. models, reducing the compute needed.
  3. Hidden spending – Some expenditures on high‑end chips are masked as “cost‑saving” techniques that make inferior hardware achieve higher performance (e.g., DeepSeek’s efficiency tricks).

• Export restrictions limit Chinese capital use – U.S. bans on advanced AI chips (Nvidia designs, TSMC manufacturing) prevent China from buying the most powerful hardware.

  • Chinese firms are pushed toward domestic alternatives (Huawei, SMIC).
  • Sanctions also block access to cutting‑edge chip‑making equipment, forcing costly work‑arounds and capping production capacity.

• Domestic demand constraints – Chinese enterprises spend < 10 % of what U.S. firms spend on IT, despite China’s GDP being two‑thirds of the U.S. (or a third larger in PPP terms). This throttles revenue prospects for AI providers, curbing their willingness to invest heavily.

• Strategic focus differs – The Chinese Communist Party emphasizes diffusing AI across the economy, not pursuing a race toward artificial general intelligence (AGI). Fewer than ten Chinese firms target AGI, compared with dozens of U.S. players.

• Investor attitudes – Chinese investors have historically punished over‑spending on AI, whereas U.S. investors once rewarded aggressive budgeting. This cultural difference keeps Chinese AI budgets modest.

• Potential bottlenecks for China – Despite restraint, China may face compute shortages:

  • ByteDance experiences ten‑hour processing times for some videos.
  • Alibaba Cloud, Zhipu AI, and Moonshot’s K3 have long waiting lists or quickly sell out capacity.
  • Over‑restriction could stifle growth if AI services cannot meet user demand.

Overall takeaway: China achieves comparable AI performance to the U.S. while spending a fraction of the capital by leveraging cheaper resources, model‑distillation techniques, and a strategic focus on wide‑scale diffusion rather than raw computational power. However, export bans, limited domestic chip capacity, modest corporate demand, and cautious investors together create both an efficiency advantage and a risk of under‑provisioned infrastructure.

interesting

they provide a mesh vpn.