Hacker News: elij

New comment by elij in "RTK reports token savings, but our cost benchmarks disagree"

elij — Fri, 11 Sep 2026 14:32:56 +0000

I just spawn a subagent in the cheapest range (for example flash-lite) to summarise a tool use. It's the only way that has worked based on my benchmarks and generalises well.

New comment by elij in "Prompting Is Dead in 6 Months. Andrew Ng, Stanford [video]"

elij — Tue, 08 Sep 2026 04:02:32 +0000

he's just talking about going from ReAct [1] to Ratchet [2]. It's largely understood in all tooling (langraph, workflows/goal, honcho, bot mode etc.).

[1] https://arxiv.org/abs/2210.03629 [2] https://arxiv.org/abs/2605.22148

New comment by elij in "Zero-Mem: Zero-Token Memory Operations for LLM Agents"

elij — Wed, 05 Aug 2026 07:42:49 +0000

This is actually quite easy to implement at the harness level and the NER can be way more naive because of the typical nature of LLM dialogue (programming, long running tasks etc).

New comment by elij in "An agent in 100 lines of Lisp"

elij — Sun, 12 Jul 2026 09:24:32 +0000

it is also possible to sandbox elisp at least. I allow pure functions through by default: https://github.com/elij/macher-agent/blob/main/macher-agent-... -- more an interpreter then straight eval but works as author states

New comment by elij in "Herdr: Agent multiplexer that lives in your terminal"

elij — Mon, 29 Jun 2026 14:12:16 +0000

this works so seamlessesly inside emacs out the box without a dedicated app

https://github.com/elij/macher-agent

New comment by elij in "Running local models on an M4 with 24GB memory"

elij — Mon, 11 May 2026 00:20:46 +0000

I'm using the 30b MOE model on same spec with 65k tokens as a sub agent with tooling and it absolutely writes decent code. The dense 9b I agree wasn't great.

New comment by elij in "BGP handling bug causes widespread internet routing instability"

elij — Tue, 27 May 2025 14:24:53 +0000

if you're a linux person consider a routing on host setup with FRR with /32s. As every host is a /32 network you can focus more on the aspects of BGP rather than TCP/IP.

New comment by elij in "Google releases Bard to a limited number of users in the US and UK"

elij — Wed, 22 Mar 2023 05:54:00 +0000

I asked Bard for the best Arsenal team and it was able to choose current players. It was also able to say how many goals they scored in current season as of today.

New comment by elij in "When Rust hurts"

elij — Wed, 15 Feb 2023 15:18:12 +0000

same issue with networking (specifically per packet heuristics) and even Java hasn't got that right either.

New comment by elij in "BGP on Windows Desktop"

elij — Tue, 17 Jan 2023 11:00:06 +0000

route on host (rfc5549) is the best strategy for host mobility if your fabric is designed around that idea and you embrace ECMP, stateless networking and bipartite topologies. You need to track state to do that without changing the FIB.

New comment by elij in "UK Police to get new powers to shut down protests before disruption begins"

elij — Mon, 16 Jan 2023 23:36:49 +0000

yea we did this in the UK during the 2011 riots, not technically protesting but same as this law.

New comment by elij in "UK government set to extract hospital data to Palantir without patient consent"

elij — Sat, 05 Nov 2022 23:50:50 +0000

so I believe this initiative is problematic but I feel activists are shooting themselves in the foot talking about privacy and conspiracy. The issue is lock-in -- the authority is mandating compatibility with proprietary Palantir feature sets (for example approach to federation and access control) and use at a local level (outside NHS-E) as a cost saving measure (the cost saving part means replace not dual operate 2 platforms). This is signaled strongly here and in the FDP procurement.

New comment by elij in "Liz Truss resigns as UK prime minister"

elij — Thu, 20 Oct 2022 13:04:42 +0000

these leaders are a reflection of the voting population in this and those respective countries.

turns out we don't have that many smart people in the UK

New comment by elij in "Meta Quest Pro"

elij — Wed, 12 Oct 2022 12:27:12 +0000

research has found a bigger correlation with indoor time in early years than computers

New comment by elij in "Sensisble Soccer"

elij — Tue, 11 Oct 2022 06:55:22 +0000

I still refer to SWOS as the greatest game of all time

New comment by elij in "Rejecting data demands, ExpressVPN removes VPN servers in India"

elij — Thu, 02 Jun 2022 14:49:39 +0000

interesting that they're having so much success with false RIPE/ARIN entries. Proper geolocation (as in with visiblity of most T1s) would trivially identify the origin of traffic.

New comment by elij in "Rejecting data demands, ExpressVPN removes VPN servers in India"

elij — Thu, 02 Jun 2022 14:31:13 +0000

not true -- your prefix location is a BGP tag which is appended based on where you're physically connected (which T1 carriers will do). Obviously you can get around it with an overlay network but you'll need some trivial PoP in India.

New comment by elij in "Russia Sanctions Add Shock Internet Censorship Twist for UK ISPs"

elij — Sun, 01 May 2022 12:51:22 +0000

they're also dealing with the restrictions on switching fabrics (that were shoe horned in with the 5G hardware restrictions)

New comment by elij in "Zero-Knowledge Middleboxes: enforce network policies on encrypted traffic"

elij — Sun, 17 Apr 2022 16:57:17 +0000

I like this idea and this is very much the future. I wish it wasn't out of band and these proofs could be embedded in the relatively small address space of IP headers. This also lets us take advantage of merchant silicon for branching logic.

Obviously some work is needed in the host IP stack for per packet tagging (DCSP/traffic class) rather than per socket and then flushing.

New comment by elij in "Deep learning on electronic medical records is doomed to fail"

elij — Tue, 22 Mar 2022 15:22:24 +0000

I agree with all points with respect to static EMRs including NLP efforts.

The cadence, uniformity and alignment of event stores to underlying pathways does pose an opportunity (but there's so little research and I suspect Brian didn't have access to this space). An EMR is a projection/point in time snapsot of these events (basically API calls). There's also inherent natural labels because inferences are evaluated against an actual pathway.

Specifically speaking about intra-episodic inferences -- longitudinal predictions (over several episodes -- or a patient life time) becomes wildly inaccurate. There's a lot of research to demonstrate this no matter which models are used.