๐ฆ๐ฝ๐ฒ๐ฐ๐ถ๐ณ๐๐ถ๐ป๐ด ๐ฐ๐น๐ฒ๐ฎ๐ฟ ๐ฟ๐ฒ๐พ๐๐ถ๐ฟ๐ฒ๐บ๐ฒ๐ป๐๐ ๐ถ๐ป ๐๐ ๐๐ผ๐ผ๐น๐ ๐ถ๐ ๐๐ต๐ฒ ๐ป๐ฒ๐ "๐ฐ๐ผ๐ฑ๐ถ๐ป๐ด" ๐
๐จ๐ป๐ฐ๐น๐ฒ ๐๐ผ๐ฏ predicted this years ago:
๐๐ฆ ๐ธ๐ฐ๐ถ๐ญ๐ฅ ๐ฆ๐ท๐ฆ๐ฏ๐ต๐ถ๐ข๐ญ๐ญ๐บ ๐ณ๐ฆ๐ข๐ค๐ฉ ๐ข ๐ญ๐ฆ๐ท๐ฆ๐ญ ๐ฐ๐ง ๐ข๐ฃ๐ด๐ต๐ณ๐ข๐ค๐ต๐ช๐ฐ๐ฏ ๐ธ๐ฉ๐ฆ๐ณ๐ฆ ๐ธ๐ฆ ๐ค๐ณ๐ฆ๐ข๐ต๐ฆ ๐ฑ๐ณ๐ฐ๐จ๐ณ๐ข๐ฎ๐ด ๐ฅ๐ช๐ณ๐ฆ๐ค๐ต๐ญ๐บ ๐ง๐ณ๐ฐ๐ฎ ๐ด๐ฑ๐ฆ๐ค๐ช๐ง๐ช๐ค๐ข๐ต๐ช๐ฐ๐ฏ๐ด.
With the rise of agentic tools like ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ผ๐ฑ๐ฒ, we are officially entering that era. ๐ป
Now, the common refrain is: "๐๐ณ๐ฐ๐จ๐ณ๐ข๐ฎ๐ฎ๐ฆ๐ณ๐ด ๐ธ๐ฐ๐ฏ'๐ต ๐ฃ๐ฆ ๐ฏ๐ฆ๐ฆ๐ฅ๐ฆ๐ฅ ๐ฃ๐ฆ๐ค๐ข๐ถ๐ด๐ฆ ๐ค๐ฐ๐ฅ๐ฆ ๐ช๐ด ๐ฏ๐ฐ ๐ญ๐ฐ๐ฏ๐จ๐ฆ๐ณ ๐ฏ๐ฆ๐ค๐ฆ๐ด๐ด๐ข๐ณ๐บ." But hereโs the reality: Code isn't going anywhere. Code represents the ultimate detail of a requirement. At a certain level, those details cannot be ignored or abstracted away by AI.
The way we use tools today requires the same rigor weโve always used: it must be accurate, formal, and so detailed that a machine can execute it without ambiguity. ๐ค
Well-specified requirements are just as formal as code, making the human factor more important than ever. We shouldn't fear this revolution; we should leverage it to move from "๐๐ฟ๐ถ๐๐ถ๐ป๐ด" to "๐ฎ๐ฟ๐ฐ๐ต๐ถ๐๐ฒ๐ฐ๐๐ถ๐ป๐ด." ๐
๐๐ผ๐ ๐๐ผ ๐ฎ๐ฑ๐ฎ๐ฝ๐:
โข ๐๐บ๐ฏ๐ฟ๐ฎ๐ฐ๐ฒ ๐๐ต๐ฒ ๐ฐ๐ต๐ฎ๐ป๐ด๐ฒ: AI is here to stay. Treat tools like Claude Code as your "staff engineer" while you remain the Chief Architect.
โข ๐๐ผ๐ฐ๐๐ ๐ผ๐ป ๐๐ป๐๐ฒ๐ป๐: Shift your energy from syntax to the architectural patterns and logic that govern the system. ๐๏ธ
โข ๐๐ผ ๐ฏ๐ฎ๐ฐ๐ธ ๐๐ผ ๐ณ๐๐ป๐ฑ๐ฎ๐บ๐ฒ๐ป๐๐ฎ๐น๐: You need strong criteria to review, audit, and catch hallucinations in AI-generated code. If you don't know the fundamentals, you're just debugging code you didn't write. ๐
The terminal is becoming a place for conversation and intent, not just commands. Let's build better, faster.
Are we moving toward a future where "๐๐ณ๐ฐ๐ฎ๐ฑ๐ต ๐๐ฏ๐จ๐ช๐ฏ๐ฆ๐ฆ๐ณ๐ช๐ฏ๐จ" is just another name for "๐๐ฆ๐ฒ๐ถ๐ช๐ณ๐ฆ๐ฎ๐ฆ๐ฏ๐ต๐ด ๐๐ฏ๐จ๐ช๐ฏ๐ฆ๐ฆ๐ณ๐ช๐ฏ๐จ"?
My daughter's meltdowns teach me something about emotions.
Her crying is real. The pain is real. She just doesn't have the tools to handle it yet.
Once it passes, relief floods in. She's fine.
That's not spoiling. That's what emotional development looks like. She's learning to move through feelings, not avoid them.
Creating skills can be hard sometimes when you try to explain your flows, but Claude has now added Record a skill.
Record your screen while you do a task, talk through it as you go, and Claude turns it into a skill it can run again
Two ๐บ๐ผ๐ป๐๐ตs ๐ฎ๐ด๐ผ, ๐ ๐ฐ๐ฟ๐ผ๐๐๐ฒ๐ฑ ๐ฎ๐ป ๐ฒ๐ป๐๐ถ๐ฟ๐ฒ ๐ฐ๐ผ๐ป๐๐ถ๐ป๐ฒ๐ป๐ ๐๐ผ ๐บ๐ฒ๐ฒ๐ ๐๐ต๐ฒ ๐ฝ๐ฒ๐ผ๐ฝ๐น๐ฒ ๐ ๐๐ฎ๐น๐ธ ๐๐ผ ๐ฒ๐๐ฒ๐ฟ๐ ๐ฑ๐ฎ๐. ๐๐ ๐๐ฎ๐ ๐๐ผ๐ฟ๐๐ต ๐ฒ๐๐ฒ๐ฟ๐ ๐ต๐ผ๐๐ฟ.
We gathered in Madrid ๐ช๐ธ โ not just for the meetings, but for something harder to schedule: real human connection.
The mission was clear: set the right foundations for whatโs coming in Deel Core over the next months. But what Iโll actually remember is finally putting faces to the names behind every Slack message, every production incident, every โIโve got youโ moment when things get tough.
These are the people building the future of global hiring โ with an agentic AI mindset and the ambition to create something truly generational.
The energy in that room was something else. And it reminded me why the work matters.
"Manager roles are dead."
Translation: "We didn't think about how to evolve them with AI."
Smart orgs are discovering: a manager with a small team + AI tools becomes an IC againโbut better. They code, design, write, lead. All at once.
That role isn't dead. It's just been reimagined.
If your business isnโt leveraging AI yet, youโre already falling behind.
AI isnโt just hypeโitโs operational leverage:
โข Reclaim hours, days, and weeks of team bandwidth
โข Automate the manual, repetitive tasks choking your flow
โข Direct resources toward high-impact, strategic work
You already know where your bottlenecks are. Take advantage of the current AI market competition and lock in a great deal for your company.
You gotta be really critic about the prompts you make, donโt get comfortable giving poor context, references, constraints.
Every new good prompt will give you better results + clear statements on what you want as output.
2026 World Cup = most tech-loaded eve
Smart ball tracks contact 500x/sec
All 1,248 players body-scanned into 3D avatars
Offside precision down to 10cm
Instant audio alerts to refsโ earpieces
Body cams on every match
AI analyzing 100M+ data points
I could never imagine how the industry was going to switch 3 years ago.
I recall reading about prompt engineering, some folks said such term shouldnโt exist.
I was mostly reading it for ChatGPT usage
I couldnโt even grasp the idea of the coding assistant we would have now.
@MarioNawfal You can have the best player in the world. But if the team doesn't share vision, culture, and values, you get infighting instead of execution.
You can't hire your way out of poor culture.
A team of rockstars without a shared vision will lose to a disciplined team every time.
Messi's last push couldn't overcome the team's immaturity.
Alignment beats talent. Always.
KFC anunciรณ finalmente fecha de inauguraciรณn. 24 de Agosto
Me gusta ver competencias, porque es buena, quiero ver como le van a las empresas locales, algunas de ellas ya son buenas y conocidas.
People using agent harnesses are the most beneficial from this AI competition, they could be shipping products really fast, by switching models.
They must have a pretty good setup though and not ending up in a big Frankenstein AI built product ๐
Building something impactful at @deel ? Our CEO just raised the stakes. ๐
Pitch how your project drives massive value for the company or our clients, and weโll double (2x) your Claude token credit limit.๐คฏ
Time to ship ๐
Frontend Code Arena leaderboard:
Kimi-K3: 1,679
Claude Fable 5: 1,631
GPT-5.6 Sol: 1,618
48 points ahead. The gap is real but not insurmountable.
Full rankings at https://t.co/y9Jtct47JC
Result: Claude executes the entire task, soup to nuts.
You review at the end, not in the middle.
This is how you get autonomous coding.
Not by hopingโby structuring the request so interruption becomes unnecessary.
How to enable true auto-mode in Claude Code
The mistake: treating it like a chat.
Ask โ answer โ ask โ answer.
The fix: front-load everything. Let it run.
Here's the pattern:
Pack the initial prompt with:
- Full ticket context (not sparse)
- Reference files & functions (not "figure it out")
- Hard constraints (not suggestions)
- The plan you want it to follow (not open-ended)
This is the inverse of chat UXโmore upfront work = zero-interruption mode.