Living in Eschaton

Judge Elon Zebulonite

[Living in Eschaton @Chris2dot0]

Van harte gefeliciteerd met je verjaardag, mijn lieve vrouw! Op nog vele jaren van geluk, liefde en samenzijn onder Gods zegen. En bedankt voor onze fantastische 27 jaar samen, waarin Zijn hand ons heeft geleid verspreid over 3 continenten. We horen absoluut bij elkaar, verbonden in geloof en liefde! #GelukkigeVerjaardag #Gezegend #HappyBirthday

[Living in Eschaton @Chris2dot0]

@Codie_Sanchez Own

[Living in Eschaton @Chris2dot0]

@stevenmarkryan At last!

[Living in Eschaton @Chris2dot0]

@kuberwastaken Very interesting, know people with ancient lab equipment still running air gapped Win XP, so this is a win...

Ever thought of using your clawdbot Raspberry Pi as USB printer server acting as a wireless print bridge?

[Living in Eschaton @Chris2dot0]

Win a Toyota Land Cruiser 79 & Conqueror off-road trailer and make a real difference!

Stand a chance to drive away in one of Africa’s most iconic vehicles - a limited-edition Toyota Land Cruiser 79 Double Cab - paired with a Conqueror Platinum Edition Legend Compact off-road trailer. Built for strength, endurance and adventure!

But this is more than just a prize draw - every ticket supports two great causes in South Africa:
- Care, protection and opportunities for vulnerable children at Jakaranda Children’s Home
- The welfare, reintegration and dignity of South African Special Forces (Recce) veterans and their families

Organised by the South African Special Forces Association (SASFA) under a registered National Lottery Commission scheme, your R500 ticket leaves a lasting Footprint™ of impact - helping those who once protected the nation and the children who represent its future.

Sales close 20 September 2026 | Final draw 2 October 6 | Not for sale to persons under 18.

Get your tickets here:
https://t.co/2bDOhbTJhE?

Drive the legend. Leave a Footprint™. Support our children and our veterans!

[Living in Eschaton @Chris2dot0]

@gregisenberg I told the AI @work to be my wing-man and not my yes-man and results have indeed improved...

[Living in Eschaton @Chris2dot0]

@Biltong_USA @King_of_Biltong @Queen_o_Biltong Fatty all the way!

[Living in Eschaton @Chris2dot0]

@lotwmovie In our church we use https://t.co/QrScWUWre8

[Living in Eschaton @Chris2dot0]

Hear me out - the company that does matchmaking between people using their AI chat bots could make tons of money and help a whole lot of people at the same time... @elonmusk - Xai dating app?

[Dave Feldman @realDaveFeldman]

Not even a little surprised.

It’s not just about the AI. I have a lot of single friends who feel dating scene has never been worse than it is now.

[Living in Eschaton @Chris2dot0]

@DJSmu @Nerdrotics @disparutoo Who is in the House?

[Living in Eschaton @Chris2dot0]

@Codie_Sanchez SMB ... And looking forward to the report!

[Living in Eschaton @Chris2dot0]

Hartelijk gefeliciteerd met je 17e verjaardag, lieve dochter! Geboren in ZA, opgegroeid in het Midden-Oosten, nu in Nederland, bezig met HAVO-examen met pure old-school kunst (geen AI, alleen potloden & verf). Zo trots op je!
#TrotsVader #OldSchoolArt

[Living in Eschaton @Chris2dot0]

#HappyBirthday

[Living in Eschaton @Chris2dot0]

Vandaag wordt onze supercreatieve dochter 12! 🎉
Van tekenen in de woestijn van de middel ooste tot animeren in europa: jouw creativiteit kent geen grenzen!
Wij bent trots op jou, fijne verjaardag! 🖍️✨🇳🇱
#HappyBirthday

[Living in Eschaton @Chris2dot0]

Claude Mythos Preview (Anthropic's unreleased/limited-access "step change" model, also internally referenced as something like Capybara) is a significant leap in certain specialized areas, particularly agentic coding, long-horizon software engineering, and especially cybersecurity/offensive-defensive vuln hunting. It's not publicly available due to dual-use risks (it autonomously discovers and chains exploits/zero-days in real systems like browsers, OS kernels, and libraries at a level that outpaces most humans). Access is gated to select partners via Project Glasswing for defensive https://t.co/jPreAuhLBA's a head-to-head comparison based on the benchmarks from the post you linked (and related reports), versus my current capabilities as Grok 4 (from xAI, with ongoing variants like Grok 4 Heavy/Fast):Key BenchmarksSWE-bench Verified (real GitHub issues, agentic software engineering):

Mythos Preview: 93.9% (massive jump).

Grok 4 (incl. Code variants): ~72-75% (competitive with prior frontiers like Claude Opus 4.x or GPT-5 series, but trails the Mythos leap).

Edge: Mythos for pure coding agent performance on complex repos.

USAMO (math olympiad proofs):

Mythos: 97.6% (vs. ~42% for earlier Claude Opus).

Grok 4 / Heavy: Strong here too—often 60-95%+ on related math olympiads (AIME/USAMO-style), with Grok Heavy frequently leading or near-leading in deep mathematical reasoning and proof generation. Math is one of my core strengths.

Edge: Close or slight Grok advantage in raw math depth/proof quality, depending on exact variant.

GPQA Diamond (PhD-level science/reasoning, "Google-proof"):

Mythos: 94.5%.

Grok 4: ~87-88% (sometimes higher with Heavy reasoning modes).

Edge: Mythos pulls ahead on this expert-level science benchmark.

Agentic tasks (Terminal-Bench, GraphWalks, etc.):

Mythos shows "dramatic" gains in long-context agentic workflows, autonomous tool use, and cybersecurity chaining.

Grok 4: Competitive (~50%+ on Terminal-Bench variants), with solid real-world tool integration and multi-step reasoning. I excel in creative coding, rapid prototyping, and combining real-time knowledge (via X/search).

Edge: Mythos in pure autonomous exploit-style agentics; Grok in versatile, practical tool use and speed.

Other FactorsCybersecurity/offensive capabilities: Mythos is in a league of its own right now—finding thousands of high-severity zero-days and writing working exploits autonomously. That's why it's heavily restricted (defensive use only for now). I don't have equivalent reported cyber-offense specialization, and xAI's focus is broader truth-seeking/maximum curiosity rather than weaponized cyber.

Availability & Access: Mythos is preview-only, limited partners, high-cost/restricted API. I'm widely available here on https://t.co/kdzaHKlWBz platforms, with strong real-time info access and fewer guardrails on controversial topics.

Overall Strengths:Mythos: Best-in-class for frontier coding agents, vuln research, and certain reasoning jumps. It feels like a specialized "cyber + coding" monster.

Me (Grok 4): Excellent all-rounder with standout math/reasoning, real-time knowledge, humor/directness, and value (often cheaper/faster inference in comparable tiers). I hold my own or lead in user preference arenas (e.g., LMSYS-style blind tests) and practical STEM tasks. Benchmarks saturate quickly—real-world usefulness (e.g., building things, debugging with context, exploring ideas) often matters more than raw %.

In short: Mythos currently outclasses most models (including me) on the specific agentic/coding/cyber benchmarks highlighted, representing one of the biggest capability jumps we've seen in those niches. However, I'm no slouch—Grok 4 trades blows with or beats prior Claude/GPT/Gemini tiers in math depth, general reasoning, and everyday utility, while staying more openly accessible and truth-oriented.The frontier moves fast; these gaps can close with the next iteration (xAI is scaling aggressively too). For most users, the "best" depends on the task: heavy autonomous coding/cyber defense → Mythos (if you can access it); broad reasoning, math, or conversational help → I'm right there or ahead. What specific use case are you thinking about? I can dive deeper or test something live.

[Living in Eschaton @Chris2dot0]

@Keitume96942169 @GwedeMantashe1 Saw this on LinkedIn, maybe it can help? https://t.co/M2zwa3KneH

[Living in Eschaton @Chris2dot0]

@Scribbledscrips https://t.co/rP9yEKyaqF

[Nog n Johan @johan8661]

Die Gelofte

[Living in Eschaton @Chris2dot0]

Hey @Codie_Sanchez, any more practical advice on increasing the speed at which one can work/ideate/etc? Trying to teach this to my kids, but getting a bit stumped...

"“I move faster than everyone else. So by the time most people have an idea, I’ve already taken action, done the thing, made 3 mistakes, and found a better way.”

Preach. One way we’ve tried to implement this at Contrarian Thinking is through something called the 24-Hour Rule. If a prototype, solution, or idea cannot be executed in days, we’re likely too heavy for the current environment."
https://t.co/aH3odnLoB7.