link-

I have started preparing a hybrid (pre-recorded + live sessions) course on practical system design that will be shipped on maven The course is designed for engineers with at least 3 years of experience, & these are the expected learning outcomes: TLDR; I want to teach the bulk of what staff+ engineers learn in a decade ⏺ Design fault-tolerant distributed systems & learn the tradeoffs of consistency and handling high-throughput workloads ⏺ Develop zero-downtime migration strategies for large-scale distributed systems ⏺ Learn proven design patterns applied in designing systems at big tech companies ⏺ Protect your systems with abuse prevention patterns ⏺ Learn how to observe & operate distributed systems in production ⏺ Level up from Junior+ to Senior and Senior to Staff / Architect ⏺ Bonus: conquer the system design interview with in-depth knowledge & practical insights You can now pre-signup for the course (link in the first comment because big boss Elon wants it that way 🙄)

@MegaBasedChad Fake news

GitHub developers don't even do post mortems for the outages anymore, they just look at you like this

It’s the brain rot of educational systems. “Everything must have a single correct answer, quickly figuring out the correct answer equals superior intelligence.” This mental programming is obviously ridiculous, and not practical. Some spend an entire lifetime stuck in this way of reasoning. The internet just amplifies this.

Its crazy how many people think slowly changing your mind and testing your theories is bad or makes you dumb.

@serengil @sspaeti @CommonSenseMars @github I think the issue is resolved now, you’ll receive communication about this shortly

@sspaeti @CommonSenseMars @github @BassemDy I'm having exactly same problem with Simon. I appreciate if you can help me, too. My github has been inaccessible for 6 days, and community cannot use my open source projects.

My GitHub user got flagged without any notification and is now unavailable to the public, and I'm not sure what the issue is. I found out by accident, when a GH action did not run any longer.. Luckily, I can still push and see my repos. But anyone else can't, and gets a 404. Thank you, GitHub. I can't imagine doing anything differently or in violation of the guidelines. Has anyone had a similar experience? This is really frustrating and shows how dependent one is, and why I should be more self-hosting to be independent. Your account can be locked at any time without notice for no reason. What's the best way forward? Anyone using Codeberg, Tangled? I have my self-hosted Gitea, but I'm not sure if I want to move everything there 🤔? Moving makes even more sense when you consider they are training on my private repos and my data, including my book and others'.

@sspaeti Simon, I received word from the team that your account is reinstated. I don't have more details as to why your account was flagged. You can follow-up on that in the support ticket

@mehd_io @cassidoo let me check with the team

@BassemDy, @cassidoo do you know anyone who can help ? Simon is one of the biggest voices in data engineering writing. I've worked with him personally for years and I can't imagine this is warranted. Would be a shame to lose his work.

Sure.. becomes a shit economy, liquidity concentration in the AI bubble, and someone starting WW3 are all great for business

Noo…. who would have thought…

@ahmedelgabri @ahmdelemam Great writeup!

So I wrote about this pattern, which I have been using for years now. gabri.me/blog/git-workt…

I have been using git worktrees for years, before agentic coding became trendy. I have been getting many questions from friends and colleagues on how I work with them.

@wesbos I never thought we’d have to worry about scaling that damned contributions graph…

Bros GitHub contribution graph about to go neon green

@kitarp29 We listen and we do not judge 😏

AI.

@moaz_eldfrawy 😄 yeah I remember that one, it was pretty dumb..

@BassemDy me.pcmag.com/en/ai/31089/vi… At least it's not worse than this

@ahmdelemam 😂

So.. who's going to package @openclaw in a mac mini and sell it as the ultimate personal assistant product?

Clawdbot is fascinating! Took me a few hours to get it setup on an isolated VM on an untrusted VLAN, with a shared folder from my NAS for memory, all running in my small homelab. It even setup whisper to transcribe my voice notes and respond to my voice prompts! If this thing is capable of running Nvidia’s personaplex for full-duplex speed to speed conversations? Game over!

What my LLM says. Publication Assessment Journal: Frontiers in Nutrition • Impact Factor: ~4.0 (mid-tier) • Credibility: Mixed. Frontiers is a legitimate peer-reviewed publisher, but has faced criticism for sometimes lax review standards. It's on the "predatory-adjacent" watchlist for some academics, though not outright predatory. The journal is indexed in PubMed/MEDLINE, which is a positive sign. • Open Access: Yes (author-pays model, which can incentivize volume over quality) Verdict: Acceptable but not top-tier. I'd trust it more than a no-name journal, less than JAMA/Lancet/NEJM. Author Assessment Authors: Junhui Jiang, Hu Zhao, Jiong Chen, Junhao Du, Weixiang Ni, Baohua Zheng, Junhong Wu, and Chunhong Xiao Affiliations: • Fuzong Clinical Medical College of Fujian Medical University, Fuzhou, China • Department of General Surgery, The 900th Hospital of Joint Logistic Support Force, PLA, Fuzhou, China Assessment: • These are Chinese military/academic hospital researchers • No major red flags, but not internationally recognized names in nutrition or oncology • Corresponding author (Chunhong Xiao) uses a personal email domain — minor concern • Limited ability to verify their publication track record without database access Verdict: Credible academics, but not leading experts in the field. Study Design & Findings What they did: • Analyzed NHANES data (2007-2018) — 25,879 U.S. adults • Estimated dietary creatine intake from meat/fish consumption • Looked for association with self-reported cancer diagnosis Key Findings: • Each standard deviation increase in creatine intake → 5% lower cancer risk • Strongest effect in males (7% reduction) and overweight individuals (8% reduction) • Older adults (66-80) showed stronger protective association • Underweight individuals showed OPPOSITE effect — higher creatine = higher cancer risk Reliability Assessment Strengths: • Large sample size (25,879) • Used NHANES — a well-respected, nationally representative dataset • Adjusted for many confounders (smoking, alcohol, BMI, etc.) • Transparent methodology Weaknesses: 1. Cross-sectional, not causal — They can only show association, not that creatine prevents cancer 2. Self-reported cancer — Subject to recall bias 3. Creatine intake was estimated, not measured — they assumed 0.11g/oz for all meat 4. Reverse causation possible — Sick people may eat less meat 5. Confounding by healthy diet — People who eat more meat in "normal" amounts may have better overall diets 6. Effect size is tiny — 5% risk reduction is barely clinically meaningful 7. Underweight paradox — The opposite finding in underweight people suggests the relationship is complex/confounded Bottom Line Is this credible? Moderately. It's real research, properly conducted, but: • It's observational, not experimental • The effect is small • The journal is decent but not elite • Don't change your diet based on this alone What it actually shows: People who eat moderate amounts of meat may have slightly lower cancer prevalence. That's it. It doesn't prove creatine prevents cancer.

It just doesn't stop 😂

Why is this regurgitation of what's already open sourced in abundance so impressive? I would be more impressed if this was solving an entirely new problem by remixing existing knowledge. Recreating a utility using what it was trained on seems like a massive waste of cycles.

I’m not sure about the usefulness of agents dropping as your experience with a tech stack increases. If anything, today, the usefulness grows with your experience. At least anecdotally speaking. I’m delegating entire problems to agents and it’s solving them at least as good as I would but Nx faster.

@BassemDy My experience so far: For easy and predictable problems, I can see AI speeding up things during the coding phase. However, the better you are at your stack and typing, the less useful AI is. For hard problems, the bottleneck is figuring out what to do, not coding!

Writing code has never ever been the bottleneck. If I unleash 10 agents running in parallel 24/7 solving well defined problems, AND deploying changes, handling operational problems, troubleshooting alerts, mitigating incidents, I would still have a backlog that takes years to get through and define well enough for an agent to take a shot at

@Hesamation @AnthropicAI @bcherny The same cycle repeats again and again since the days of vi

if anthropic released an IDE with their own claude integration and a lighter design and less bloat than vscode… the software industry would collectively lose its fucking mind overnight. just throwing this out there @AnthropicAI @bcherny