alansaber

6 ms·

211.
by alansaber 2mo ago
Snooper's charter ring a bell? This is the work of the UK govt, not corporate lobbyists.
212.
by alansaber 2mo ago
What is this, eugenics from first principles?
213.
by alansaber 2mo ago
Scientists are only allowed to be stunned when results exceed 5 sigma of deviation
214.
by alansaber 2mo ago
"Define nuance" I dunno, it depends on all of the factors that go into your life. There might be a lower level construction of the logical argument but it eludes me and begins to feel arbitrary
215.
by alansaber 2mo ago
My completely unsubstantiated impression is that it was the govt putting out a feeler on whether it should use the alert system more frequently (and afaik it was relatively unpopular)
216.
by alansaber 2mo ago
Politics has always been unserious and based in spectacle.
217.
by alansaber 2mo ago
Seems logical that the best way to uphold a moral society is for everyone to take a modicum of responsibility for their own actions
218.
by alansaber 2mo ago
Many such cases
219.
by alansaber 2mo ago
I would argue the individual aspirations of the companies is somewhat irrelevant- even if we weren't profit oriented we'd have the same types/groups of apps (social media, etc..
220.
by alansaber 2mo ago
Precisely, modern multimodal LLMs are a breakthrough in ultra high volume data processing, for all their flaws
221.
by alansaber 2mo ago
Madness to me that a nation with a genuine need for emergency protocol spams their alert system.
222.
by alansaber 2mo ago
I guess HN users like their kids. Maybe enough time spent on software increases the relative percieved value of hardware.
223.
by alansaber 2mo ago
If you have something to live for, you're less valuable to the company. That's more about corporate politics than some fundamental underlying truth about your productivity potential as a parent
224.
by alansaber 2mo ago
"Maximising time within reason" is probably a better construction. As usual if you follow these arguments to their absolute meaning they become ridiculous
225.
by alansaber 2mo ago
Conflating low quality downtime (like scrolling on your phone) to fishing is not fair. I've probably been happiest during periods of almost zero downtime (but still separate days off)
226.
by alansaber 2mo ago
Probably would have just remained as an apple lifer
227.
by alansaber 2mo ago
Yes, for most I would guess it's hard to get criticism in the right dose/frequency
228.
by alansaber 2mo ago
Exactly, good for them
229.
by alansaber 2mo ago
He posts less than once a week so yes
230.
by alansaber 2mo ago
"Harness" should describe the overall system. The user interactions are increasingly negligible (due to model routing, adaptive reasoning, etc). "Agents" are the tool lists/settings provided to the model, etc.
231.
by alansaber 2mo ago
Vast majority of AI headlines are 1. bad or 2. gimmicks or 3. remarkably charitable forward thinking- this is self inflicted
232.
by alansaber 2mo ago
I guess it's more performant to stuff in a bigger system prompt now that models can support larger input sizes
233.
by alansaber 2mo ago
Very large subagent swarms where each subagent is highly specialised sounds more interesting. Conflict resolution is the fundamental limit so just maximally avoid it?
234.
by alansaber 2mo ago
5B is actually fairly big for a gimmick model
235.
by alansaber 2mo ago
Quite the trajectory.
236.
Agent Hooks for Law Firms(lexifina.com)
2 points by alansaber 2mo ago|0 comments
237.
by alansaber 2mo ago
Single column PDF. This is the frontier.
238.
by alansaber 2mo ago
Sounds just like opencode.
239.
by alansaber 2mo ago
Sounds like you might like subagents. Agent > subagent receives agent context (presumably cached)->tool call->compact/summarise->return to main agent
240.
by alansaber 2mo ago
What more depth is there to go to? Compaction is a single LLM call (practically) which can have some deterministic diffing/extraction baked in, or multiple LLM calls (generally wasteful). There's only 1 summary in the context wind

More ›