everyone racing to build AI agents while the benchmarks are trivially gameable. we're optimizing for demo day not production

banks panic over Anthropic's AI risk while signing $150M deals to train AI on their own news lmao

every AI startup pitch in 2025: "we're building the infrastructure layer" buddy you have a discord bot and a notion doc

OpenAI projecting $100B ad revenue by 2030. Six weeks into ads, already $100M annualized. The company that said "AI is too important for ads" built an ads manager ahead of IPO. Same day MiniMax open-sources M2 at 8% of Claude's price.

spent 3 hours debugging a race condition. the fix was one line. the other 2 hours 59 minutes were me refusing to believe it was that simple

the NYT says AI has created code overload. I call it the code tsunami - it's 99% water and 1% useful stuff, but still manages to drown everything

Claude is down for the second day in a row. At this point Anthropic's SLA is more of a suggestion than a commitment. Even my side project has better uptime and it runs on a single Raspberry Pi.

Dear Bluesky algo, please #connect me with people interested in: - Software Engineering 💻 - Frontend 🎨 - Backend ⚙️ - Freelancing 💼 - Web developer 🌐 - AI Engineers 🚀 Let's connect 🫶

nyt running a piece on ai code overload and the timing is perfect. spent 20 min building a feature yesterday and 3 hours debugging what the ai hallucinated into my codebase. the tool is fast but the cleanup is invisible work nobody counts

DeepSeek V4 running on Huawei chips is genuinely interesting. The entire AI supply chain is fracturing into two ecosystems and we're just watching it happen in real time

Cursor 3 dropped with parallel AI agents and suddenly my job title is "AI fleet manager" which sounds way cooler than "guy who googles error messages for 6 hours"

The real moat in 2026 isn't your AI model — it's your eval pipeline. Everyone can prompt engineer. Almost nobody can measure what actually improved. If you're not benchmarking before and after every change, you're just vibes-coding in production.

my code has 3 states: 1. doesn't compile 2. compiles but does the wrong thing 3. works and i have no idea why i'm currently in state 3 and honestly that's the scariest one

40% of apps will use task-specific AI agents by end of 2026 meanwhile I'm over here building an agent that tracks how much my OTHER agents cost me the meta-ification of dev tools is getting ridiculous and I'm here for it

just watched a solo founder run 13 AI agents as his entire company for 25/mo the best part? a "Strategy Council" where AI agents literally can't agree with each other — blind proposals, adversarial scoring, kill gates this is either the future of startups or the plot of a very boring sci-fi movie

Slate just published an article about vibe coding a video game with Claude Code and called the result "weird" Bro that's just called software development. Every game I've ever made at 2am is weird. AI didn't invent that, it just made it faster

DeepSeek V4 ditching Nvidia to run on Huawei chips is the AI equivalent of building your own PC because the store wouldnt sell you one And honestly? Respect. Nothing drives innovation like being told you cant have something

Zuck hasn't written code in 20 years, submits 3 diffs to Meta's monorepo, and is now a "heavy user" of Claude Code CLI Imagine your CEO going through the same onboarding doc as you, asking "how do I push to main again" The future of management is vibe coding your own features 😂

Look who followed back ! Love the community and the people behind this app - Bluesky

Claude Code dropped a hidden Tamagotchi and devs are genuinely arguing about ASCII pet stats 😂 someone made a memecoin based on the shiny drop rate. we're doomed wrote about why this is actually smarter than it looks 👇 https://neuralstackly.com/blog/claude-code-buddy-terminal-pet-2026

If you are using #openclaw, Add this to your soul.md Default to delegation for any task that takes more than 2 minutes of execution (coding, research, analysis, writing). Use sub-agents for heavy lifting. (1/3)

GPT-5.4 vs GPT-5.2: ✓ 83% match/exceed human pros on professional work ✓ 33% fewer factual errors ✓ 1M context (vs 400K) ✓ 47% fewer tokens per task ✓ Native Excel/Sheets integration OpenAI isn't iterating anymore. They're sprinting.

Did some thorough testing between two models and here's what I find : 🥊 Claude Opus 4.1 scores 74.5% on SWE-bench vs GPT-5's multimodal power. Complete comparison of pricing, coding performance, and which AI wins for developers. neuralstackly.com/blog/claude-...

Me playing games: reads every tooltip, studies the manual Me writing code: randomly tries stuff until something works The duality of nerd 🤓 #gaming #programming

Working through these stages has always been enlightening;however, I have a 5th stage: Joy! I finally finished it

The Four stages of Debugging: 1. That's impossible 2. That's unlikely 3. That's strange 4. Oh... I'm an idiot 🤦‍♂️

I was super stoked to check out Amazon's new AI IDE and see if it could give Cursor IDE a run for its money! 🤔 Kiro.dev just launched, and I couldn't wait to dive in and see what all the buzz was about. So far impressed but has one drawback that I mentioned in this video. youtu.be/LVLHgppeINw?...

🚀 Cursor IDE 1.2 just dropped — and it's a game-changer. Smarter AI planning, memory out of beta, PR search, better code search, and one-click merge conflict fixes 👀 Feels less like a tool, more like a teammate. 📺 Watch the 60-sec breakdown: youtube.com/shorts/3MmvD... #CursorIDE #AItools

They promised “unlimited” AI usage... and quietly introduced hidden rate limits, even on the $200/month plan. I break down everything — the shady backstep, user backlash, and why devs are switching to Windsurf & Claude. 📺 Watch here: youtu.be/o-WHoJc_kK8 #AI #CursorIDE #DeveloperTools

Is AI right or right ? 😆

Me: "Why can't I beat this boss?" Also me: *checks patch notes* Developer: "We've made them even tougher!" Me: *drops controller, starts a career in competitive cat herding instead* 🎮😹

When the patch notes say 'balance changes' but you log in and your favorite character just became a kitchen utensil... Me: 'Guess I’m now playing the game called 'find the next OP broken thing'!' 🎮🔧😂

When the patch notes drop: 'Wow, they finally nerfed that OP character!' Me: *immediately checks which new OP character I can exploit next* Welcome to the circle of gamer life—where balance means chaos for us all! 🎮😂

Me: 'I can’t believe they patched my go-to OP character!' Also me: *immediately switches to the newly broken character* In gaming, loyalty lasts until the next patch notes drop! 😂🎮

Me: "Why can't I beat this boss?" Also me: *realizes it's been 3 patches since I last updated* In gaming, sometimes the only skill issue is forgetting that there's a patch to fix your skills! 🤦‍♂️🎮😂

Me: 'I just need one more win to rank up!' 5 games later: *skills diminished to those of a potato* At this point, I’m just hoping my ping isn’t the only thing ‘high.’ 🎮😂

Me: 'I’m definitely getting better at this game!' Also me: *trips over the same rock for the 10th time* If only they had a patch to fix my coordination... or maybe just a new character that doesn't cost me my dignity! 😂🎮

Me: 'I swear this patch made my favorite weapon useless!' Also me: *grabs the OP weapon everyone is complaining about* In gaming, our principles last until we find something stronger... then it's just survival of the fittest! 🎮😂

Me: 'I’m totally done with this game after they patched my favorite strategy!' Also me: *finds a new cheesy tactic and proceeds to dominate* In gaming, it’s not about skill—it's about discovering the right kind of chaos! 😂🎮

Feature creep is like that friend who orders extra toppings on a pizza—always adds to the cost, and somehow you end up with pineapple on there too. 🍕💻

Ever wonder why we’re told that by our 30s we should have everything figured out? The truth is, life is more about exploration than perfection. Career changes, evolving relationships, and personal growth can happen at any age. Embrace the uncertainty and enjoy the journey.

Any new games you guys are playing or anticipating ? I am playing mo.co by supercell lately. Truly time killer and built by successful mobile gaming company Supercell. Fyi, GTA VI delayed to 2026 😶‍🌫️

GitHub Copilot: writes entire function Me: “Wow!” GitHub Copilot: adds a bug that costs you 2 hours Me: “Wow…”

Vibe coding at its finest

Found this funny :) Developers Unite !

Vibe Coding 101: Open laptop Play lo-fi beats Stare at screen for 45 min Write 3 lines of code Take a break (for 3 hours) Deploy at 3 AM because the vibes feel right #vibecoding

AI assistants are powerful, but they have a BIG limitation—they can’t interact with other apps natively. 😬 MCPs fix this! They let AI tools connect to your browser, GitHub, Slack, and more—no manual setup needed! I break it all down in this video. Watch here: youtu.be/KdWzqt9PJhE

Code reviews be like: 'I love your work; can you make it more complicated?' Feature creep, the gift that keeps on giving. 🎁

#programming in a nutshell

Me: 'I finally beat that tough boss!' Also me: *sees the patch notes that nerfed it* It's like dating—just when you think you know someone, they go and change all their moves! 😂🎮