Сподіваюся робочий тиждень у всіх почався добре!
Ось смішне з твітеру:
https://x.com/0xgaut/status/2011099638817018345

Але, сам вже з півроку користуюся cursor, і звик! Задачі типу «перепиши все, щоб було добре» це поки занадто, але конкретні реалізації невеличких частин коду він робить переважно добре. По моїм відчуттям економить багато часу, який був би витрачений на пошук документації і у документації, та й на знайомство з кодовою базою. Хто не спробував ще - спробуйте, може й вам буде до помочі

Це просто шикарно (з просторів лінкедіну):

During a 1:1 at Amazon my manager said:

"If you want to grow in tech, you need visibility. The best way to achieve this is to take initiative."

So I rebooted the production database without being asked.

Now I’m known across teams.

:)

Коли АІ спілкується з АІ: they are absolutely right😂

◆ THE UNIX WAY ◆
Episode 05: The Original Microservices

grep was written in 1973. awk in 1977. sed in 1974. sort, uniq, cut, wc: all before 1980. Each does one thing. Each takes text in and puts text out. Each composes with any other through a pipe.

The industry spent the last decade building vastly more complex solutions for what these tools have done all along. Then it called the pattern "Microservices." Quite.

■ The Pattern

Every microservices tutorial teaches the same principles. Unix had them in 1973:

Single responsibility: each tool does one thing. grep filters. sort sorts. uniq deduplicates. No tool tries to do two things.

API contract: stdin and stdout. Text in, text out. The universal interface. No schema registry, no versioning endpoint, no contract testing framework.

Message queue: the pipe. One process writes, the next reads. Backpressure built in. Zero configuration.

Service discovery: $PATH. The shell knows where every tool lives. No Consul, no etcd, no DNS sidecar.

Orchestration: the shell script. Sequential, parallel (xargs -P), conditional (exit codes). A few lines, not a YAML cathedral.

Observability: tee. Tap into any data stream between any two services, live, without modifying either one. No agent, no $23/host/month.

■ The Example

grep 404 access.log | cut -d' ' -f7 | sort | uniq -c | sort -rn | head -20

Six services. One pipeline. Each stateless, each composable, each replaceable. 10,000 requests analysed, top 20 error paths, sorted by frequency. Milliseconds. FreeBSD, Debian, macOS: it is already installed.

The industry sells you ELK for this. Elasticsearch alone: 4-8 GB heap. grep: 2 MB. One might call that a markup.

■ The Repackaging

Docker took cgroups and namespaces (kernel features since the 1990s) and added a daemon, a hub, and a subscription. FreeBSD jails did this in 2000, thirteen years earlier: 2-4 MB overhead vs 50 MB Alpine, no daemon between you and the kernel.

Kubernetes buried iptables, cgroups, and etcd under 12-24 GB of control plane RAM. Marvellous engineering, that: making reliable things unreliable, then selling the fix.

■ THE POINT

These tools are compiled, optimised, battle-tested for half a century. They implement the microservices pattern in its purest form: small, stateless, composable through a universal interface.

Nothing wrong with selling good solutions. But these created more problems than they solve. Larger attack surface. Hardware costs multiplied by factors the originals never required. New roles hired to maintain what needed no maintaining. Staff sent through certification programmes so they can operate licenced products that replicate what the kernel already provides.

The industry did not ignore the Unix philosophy. It repackaged it, not to make it more accessible, but to make it more billable. Everything underneath was already there.

The tools are there. They always were. Rather patiently waiting, one might add.

Зкопійовано звідси: https://www.linkedin.com/posts/vvoss_theunixway-freebsd-linux-activity-7431627385630007297-8Fdl

Пан Альтман, звичайно, хайпує і просто привертає до openAI увагу, але імхо не безпідставно. Якась доля правди в його словах, мабуть, є.

https://futurism.com/artificial-intelligence/sam-altman-thanks-programmers-over

Ні, не жарт
https://arstechnica.com/ai/2026/03/entire-claude-code-cli-source-code-leaks-thanks-to-exposed-map-file/

Репозиторій: https://github.com/nirholas/claude-code

Якщо все ще маєте трішки часу у вихідний день - можна завершити цим відео з інтерв’ю з Мартіном Клепманом:
https://m.youtube.com/watch?v=SVOrURyOu_U

Почув нещодавно про cmux - хайповий термінал, зроблений під використання усіляких АІ агентів, такий от продукт свого часу. Імхо варто спробувати - ідея з нотифікаціямм та воркспейсами уже норм сама по собі. Ділюся посиланнями.

Коротеньке демо:
https://youtu.be/i-WxO5YUTOs

Сторінка:
https://cmux.com/uk