Quick demo of JrDev - my coding agent built entirely in Python (backend Textual frontend) 🐍 Such a fun project to build! Free & Open Source Link below 👇

Opus 4.5 passes my "who is presstab" test with flying colors - the most accurate I've seen so far. This is an empty context knowledge check: no web search, no conversation history. Impressive given how obscure these facts are.

I've been quieter on X lately... honestly struggling to find the time to work on personal hobbies. Teaching Databases II has consumed most of my energy (3X weekly, in person, 2.5 hour classes!). Never realized how much prep goes into teaching when you genuinely care about student learning: - Creating engaging labs that reinforce concepts - Designing classes that blend lecture/whiteboarding with hands-on work and strategic YouTube videos. Attention spans are finite, and databases can be a brutal topic at 6pm - Grading assignments while providing meaningful feedback - Answering the constant stream of emails and Teams messages (definitely underestimated this one) - Studying edge cases to keep everything fresh in my mind It all adds up, but I'm genuinely enjoying it!

Happy to see you in my terminal Jules! x.com/simpsoka/status/197381…

Is this subsidized token era beginning to come to an end? x.com/BuiltByVibes/status/19…

Close to hitting the 1000 free requests limit today on Open Router using grok-4-fast 🙃

I ran slight mutations of the "Sonnet Fails" evaluation to test instruction adherence across models. ➡️Results on this eval set: Sonnet 4.5 improves over Sonnet 4, passing 5/8 runs vs 3/8 for Sonnet 4. 🤔 Is failing this evaluation part of the secret sauce for good agentic models? Two legendary agents, Sonnet and Codex, don't adhere to instructions well here. Yet they dominate agentic coding tasks. They seem to march to the beat of their own drum, prioritizing problem solving over rigid compliance. 🔮 The "Oracle" models (2.5 Pro, Opus 4.1, Grok-4) ace instruction following. These are all smart models, but aren't fan favorites for agentic coding. Does their precision might actually limit their ability to make creative leaps that complex coding requires? ⚡ Grok-4-fast is interesting. Perfect instruction following while maintaining speed suggests great training optimization. I use this model often (its free on Open Router) for large context summarization (ie look at a code diff and make a commit message from it). 🗣️Quick shoutout to 2.5 Flash hitting 88% ❔My Question Is there a hidden inverse correlation between some areas of instruction adherence and agentic performance? Models that interpret rather than rigidly follow instructions might actually be better at navigating complex coding workflows. This challenges what we think makes an effective AI coding assistant. At the same time though, its not like we are all looking at the Qwen models, which failed spectacularly, as legendary agents. ➡️For strict compliance tasks, pick the Oracle models. For creative problem-solving in code, Sonnet and Codex's flexibility gives them the edge despite lower eval scores.

Someone with a "job" sent me a coding challenge. Seemed weird. Uploaded it to Gemini 2.5 Pro. Worth the $0.92 I spent for the analysis. Grok-4-fast told me it was low risk 🫠

Sonnet 4.5 still fails this instruction following eval (most models pass). New DeepSeek v3.2 passes with no issue. Sonnet has problems following instructions when it has the ability to collect more context - even when directed not to. x.com/presstab_dev/status/19…

DeepSeek v3.2-exp is performing well for me so far. Reasonable time to first token, very coherent agent operations, no failed tool calls. Getting about ~25 tokens per sec. Looks promising. x.com/deepseek_ai/status/197…

This is a great take

Pretty much the only way to create something that will still work 10 years from now, with little or no maintenance. x.com/thdxr/status/197163734…

Terminal text this smooth is wild. Why haven’t the top CLI tools recruited Will yet??

Is AI a bubble? Not fully convinced yet. The "bubble" exists in one saturated corner: dev tools. Similar products, inflated valuations, very little customer loyalty. Meaningful AI adoption has barely begun in massive areas: ➕ Professional creative workflows (ie production-grade video editing, not just generation) ➕ Financial analysis (many firms remain technological dinosaurs) ➕ Dynamic, personalized gaming worlds ➕ Consumer products that simplify life, require minimal user intervention, and aren't cringe Economic viability improves daily. The shift from massive general models to more efficient, specialized ones will slash compute costs. Corrections in overheated segments? Expected. But there are many sleeping giants.

The best block explorer service crypto has ever seen. One day I just decided to build it, because why not? Hand built. 5-10 VPS instances working together to index and host the data from 30 different blockchains. Ran it at cost (like $7/month per coin). 2014 to ~2020ish

Grok 4 Fast is not on par with 2.5 Pro. Simple test: summarize my ~200k token code diff. I used GPT-5-High to judge the results. G4F: 4.8/10 - Detailed but misses required sections, breaks tone/format constraints (emoji, imperatives), and includes speculative claims. 2.5 Pro: 8.0/10 - Closely follows the requested structure with clear user impact and neutral tone; only minor nit is labeling the title instead of using a plain H1. G4F does have its place and purpose though, especially considering I paid $0 for those ~200k tokens and 2.5 Pro cost me $0.60. Web search, file search, simple modifications, etc are great use cases. x.com/ArtificialAnlys/status…

🎯 x.com/presstab_dev/status/19…

One of the few AI agent's out there with a bit of personality x.com/julesagent/status/1969…

Started using AI Studio instead of ChatGPT for non-dev AI help. It gives full control: thread forking, temp slider, top-p, delete any message. 4 tabs open right now: - Image editing - Image to text from Thai to English - Syllabus edits for Databases II - Lab creation for DBII

Beta testing Marble from @theworldlabs Gaming will evolve in this direction. Starting with choose-your-own-adventure games, then advancing toward fully customizable immersive worlds. The potential for dynamic personalized gameplay is massive. x.com/martin_casado/status/1…