Mark Kretschmann
75K posts

Mark Kretschmann
@mark_k
AI & Software Engineering | Practical insights on AI, code, fitness & e/acc • Accelerating the future
SINGULARITY Katılım Eylül 2007
691 Takip Edilen45.7K Takipçiler
Sabitlenmiş Tweet

@mark_k I think it is more you then, I am seeing both mutuals and non mutuals
English

Grok Build 0.2.100 and 0.2.101 are out from @SpaceXAI, with a strong focus on smoother multi-agent workflows and session continuity.
Most important changes:
• Discover and resume recent Claude Code, Codex, and Cursor sessions
• One-click session resume from the welcome screen
• "grok inspect" now shows effective compatibility settings
• Native refresh-rate support for high-refresh displays
• Queued messages now appear immediately and send more reliably
• Better status updates for waits, parked subagents, and long-running turns
• Web fetch now preserves truncated page content as readable artifacts
• Multiple fixes for crashes, hooks, resume hints, and terminal readability

English

@Zaug72 @beffjezos We have a bunch of group chats, but other than that, it's more about the spirit. You can put e/acc in your bio or username if you like.
This is in stark contrast to the doomers, who are highly organized with NGOs and "institutes", where a lot of money is involved.
English

@mark_k @beffjezos How do we sign up, and, what's the secret handshake?
English

The old 𝕏 algorithm fragmented online communities like e/acc, a movement that was born and built on this platform. But now I’m seeing many of my e/acc frens on the timeline again.
Effective Accelerationism, started by @beffjezos, is a movement of techno-optimists, builders, AI lovers, and everyone who believes progress should be accelerated rather than feared or suppressed.
We fight stagnation, scarcity, and doomerism. We believe technology, intelligence, energy, and human ambition can create a radically better and more abundant future.
That is e/acc. Build the future. Don’t apologize for it.

English

@dcolascione @OpenAI It's a property of the harness, which is arguably a part of the model system now. You can't use GPT-5.6 without the harness.
English

Very interesting detail revealed about GPT-5.6:
In contrast to earlier @OpenAI models, 5.6 does *not* drop the reasoning context (CoT) at each turn. Earlier models only kept the context for the current turn, but dropped everything earlier to conserve context window space.
Keeping the reasoning traces for the entire run improves thinking quality! And due to caching the overhead is not as bad as expected.

Henry Broomfield@HennersBro98
Updates to GPT-5.6 reasoning context One of the quieter but more significant GPT-5.6 updates I'm excited to have played a part in is that reasoning context is preserved across turns, giving: - better KV cache hit rate - reduces latency and cost - less repeated reasoning on long-running tasks - stronger follow-up behavior in complex environments when a task depends on prior hidden thinking
English

@ConnorTalksAI @OpenAI It definitely requires more tokens, yes.
English

@mark_k @beffjezos I just wish mankind wasn't like a quadriga in which each of the horses is trying to run in a different direction 🙄
English

@perrymetzger It's a mystery to me why that Decel is so popular.
English


@mark_k Alguien con orejas de Minnie no creo que tenga demasiado poder de convencimiento. 😅🫣
Español

@mark_k Cave people decels are probably the biggest smelliest group of losers we’ve ever seen.
English

@SanogoMari38278 Nope, Notch has been posting Anti-AI stuff for a long time
English













