intellectronica.net - Profile | ThreadSky | a Reddit-style client for Bluesky

This is the ultimate AI cluefulness test. Yes, brand, UX, & integration matter. But the model itself is the decisive factor, and the differences, even between the latest SOTA models, are meaningful. People who use models a lot, and not just for chatting, know this intuitively.

submitted 42 minutes ago • 0 comments

Now that o3 is affordable, are people using it for coding? Any interesting first impressions? (not benchmarks, first-hand vibe smelling and anecdotes plz)

submitted 19 hours ago • 0 comments

Which environment do you prefer for an Online Community? (Discord, Slack, Circle, other, .....)

submitted 20 hours ago • 0 comments

Exploring new depths of regression, in today's drumming lessons I played the timeless 90s hit Fuck You I Won't Do What You Told Me (except for the really fast part in the beginning, which is impossible). It's not a full-blown midlife crisis until I join ANTIFA.

submitted 22 hours ago • 0 comments

Excellent characterisation of the skill set and attitude for successful AI Engineering. Developers who successfully assemble a versatile skillset and continuously update it can deliver powerful products and solutions. Dogma and overspecialisation are less helpful than ever.

submitted 1 day ago • 1 comment

Unpopular opinion: @code is developed by a big tech giant, but it is also an open-source project with a thriving community and deep commitment to extensibility. Forking it is dumb and uncool and can't be justified. It's a dark pattern.

submitted 1 day ago • 1 comment

Tried o3-pro for writing. Not vibe-writing, but the sort I mostly do these days, where I feed in very detailed context and instructions. I hoped for superior performance, but so far, it's producing worse output than gemini-2.5-pro and o3 (for the same prompts / context).

submitted 1 day ago • 0 comments

Every time you're tempted to think that you have an advantage because you're using some clever AI manoeuvre, remind yourself that everyone has access to the same model and the same tools, and the new equilibrium is everyone doing exactly what you are doing.

submitted 1 day ago • 0 comments

The good news: AI adoption is already significant, and productivity improvements are noticeable. The bad news: as long as AI depends on human operators, productivity improvements remain capped at a fraction of its capacity to execute. x.com/johannes_wa...

submitted 1 day ago • 1 comment

Excited about the new model release from this or that provider? The benchmarks look wow? Impressive sample outputs? Love the license / price / geo? Just remember that none of that tells you what you really need to know - how well it works on your task. ai.intellectronica.net/structured-...

submitted 1 day ago • 0 comments

www.youtube.com/watch?v=6FP...

submitted 2 days ago • 0 comments

Career advice: if you want a ridiculously high paying job and endless employment and business opportunities in the few years, get yourself to black belt in AI Security.

submitted 2 days ago • 0 comments

So @tylercowen asked o3-pro to estimate the % of labour it could replace. The estimate and review are conservative. Cowen may claim that even if it _can_ take over 15-25% of work, it doesn't mean society will make that happen. Still, I'd suggest paying close attention. chatgpt.com/share/6848a...

submitted 2 days ago • 0 comments

Just to prove that random distribution can get very lumpy, this week I went to more live shows than I have in a decade. Tonight it was Stereolab (with Astrobal for support). It was so cool!

submitted 2 days ago • 1 comment

RIP Brian Wilson. Thanks for the vibrations. 🙏

submitted 2 days ago • 1 comment

TV caricaturises how we view professions and industries we don't know well. When I talk to people about opportunities for AI in legal, they imagine that what legal (and legal-adjacent) firms do looks like Boston Legal or Suits - every case is a completely new story.

submitted 2 days ago • 1 comment

Reminder that if you want to give o3-pro a try and don't want or can't afford the $200/month pro sub, you can access it from the Open AI playground.

submitted 2 days ago • 1 comment

Now we're talking - double o3 quota in ChatGPT. Confused? Cheat Sheet updated: gist.github.com/intellectro... x.com/kevinweil/s...

submitted 3 days ago • 1 comment

I'm always reluctant to give this as advice, but, sometimes just waiting patiently for the price to drop is a viable AI cost optimisation strategy (today, for example, we got OpenAIs o3 at an 80% discount). For some other steps you can take, see ai.intellectronica.net/cost-contro...

submitted 3 days ago • 0 comments

A quiet revolution, bubbling under the surface: the Legal Services industry is leading in AI adoption. Second only to software development. Currently more than 50% of inbound inquiries I'm getting are for legal AI, with some truly original and transformative ideas.

submitted 3 days ago • 0 comments

Loved @AarushSah_ (of @GroqInc)'s talk last night - one of quite a few great guest talks in @HamelHusain and @sh_reya's excellent evals course - exploring the relationship between AI Evals and Reliability Engineering. As we move to deploying AI at scale, evals _are_ part of SRE.

submitted 4 days ago • 0 comments

People are overly sceptical about the potential for massive productivity improvements with AI in the near future because they are not sufficiently sceptical about the flavour of AI sold to them by big tech — chatty "copilots" and "agents". Wrong paradigm, wrong predictions.

submitted 4 days ago • 0 comments

Exactly a year since @leopoldasch published Situational Awareness. Still relevant, nothing in it has been disproven, future predictions look stronger than before. situational-awareness.ai/

submitted 4 days ago • 0 comments

💡 𝙋𝙧𝙖𝙜𝙢𝙖𝙩𝙞𝙘 𝘼𝙄 𝙩𝙤𝙤𝙡𝙞𝙣𝙜 𝙧𝙤𝙖𝙙𝙢𝙖𝙥 🛠️ 1️⃣ Flagship LLM + basic SDK 2️⃣ Python/TS & Notebooks 3️⃣ Frameworks only after you’ve shipped and learnt The “adopt reluctantly” mindset builds resilient, debt-free systems that scale. Read the full guide - 🔗 in ⤵️

submitted 4 days ago • 1 comment

One phrase I particularly disliked when I worked at MegaCorp was "thrive in ambiguity" — the notion that well-adjusted employees watch passively and accept ambiguity as a given. AI doesn't just thrive in ambiguity; it butchers ambiguity.

submitted 4 days ago • 0 comments

i've seen the future, brother, it is murder

submitted 4 days ago • 1 comment

... הליצנים הם קופים, והקופים בני אדם ...

submitted 5 days ago • 0 comments

Just found out that David Thomas or Pere Ubu, is no longer with us (left, too early, a few weeks ago). Farewell, mad genius 🙏 youtu.be/-juhwf1Fa6c

submitted 5 days ago • 0 comments

MCP is the VBX of our era (if you're old enough to remember, you'll recognise the conceptual parallels)

submitted 5 days ago • 1 comment

I tried various meeting AI products, all are convenient to use, but I didn't like the output quality of most, so for meetings that matter I just record with Voice Memos, paste the recording into @googleaistuid, and use Gemini 2.5 Flash with this prompt gist.github.com/intellectro...

submitted 5 days ago • 0 comments

🌅🏥💥 www.youtube.com/watch?v=Ox5...

submitted 5 days ago • 0 comments