wtf man gpt 5.6 is absolutely rl-fried when it comes to its websearch tool. in the process of searching for graph theory papers it decided to also sneak in Netflix, Steak n Shake, a trip to Universal Studios, and *five* fucking separate dictionary lookups of the word "they"
@signulll gonna give this a "yes, and" because I think we can expect sota progress to catch up to this and what follows is a fat fucking mess of incentives to offer illicit / subsidized access to alternatives
in a completely predictable turn of events the government is (in)effectively regulating AI
which itself would be fine if step 0 wasn't accidentally imposing ID verification on everyone
i am bewildered by specifically the people in AI discourse on this platform who didn't know their $20 access to thousands of dollars of compute was subsidized
if you are solving for resource optimization for yourself it's a different question than one about incentives that modify user behavior.
user behavior driven by work during primary business hours is a much harder problem you won't easily solve without creating incentive to use a competitor in those same hours.
@thsottiaux as the literal person to have put an end to unlimited storage policies in tech, when I tell you that carbon realities escape everyone I mean it from the bottom of my heart.
also I've now spent 6 years fighting abuse at Google if you wanna sidebar!
My coding workflow is currently:
1. Spend about ten minutes writing a plan for a new feature in a markdown file as bullet points.
2. Spend about an hour with a bunch of back forths between Opus 4.6 and Codex 5.4 xhigh, asking each one to improve the markdown plan file.
3. Read the final markdown plan, which ends up as a bunch of bullet points, includes implementation details, and is usually between 200-500 lines long.
4. Ask either Claude Code or Codex to implement it.
5. Ask the other one to review the implementation.
6. Run my /deslop command on the implementation (see my pinned tweet).
7. Deploy, test, and ask an agent to fix any bugs.
@emollick fwiw, I rarely bother to reply to posts because before the sloppening there was engagement farming and ragebaiting.
like I just can't find an upside that is worth perpetuating any of that trash over
The reports of my death are greatly exaggerated.
- MCP
your user does not care about the transport layer, your buyer cares about governance (not protocols) and developers will chew on glass to achieve a goal.
ya'll are just being picky about which glass you want to chew on.
We just released Claude Code channels, which allows you to control your Claude Code session through select MCPs, starting with Telegram and Discord.
Use this to message Claude Code directly from your phone.
@stecal12@Dagnum_PI you can until you get crowded in on the tram platform - this has happened to me. the escalators fill the platform quickly and then you're crowded in too tightly to leave easily