Hacker News: manthangupta109

Ask HN: Views on Mastra's SOTA Memory?

manthangupta109 — Thu, 12 Feb 2026 17:59:17 +0000

I went through both the blog and the code for Observational Memory, and really interesting direction, and I appreciate the transparency in sharing implementation details. I do have a few thoughts on the SOTA memory claim and the broader framing.

From what I can see:

1. The implementation appears heavily tuned toward performing well on LongMemEval. That's a useful signal, but it doesn't necessarily translate to robust long-term memory behavior in production environments.

2. It feels closer to context compression/context management than a durable long-term agent memory system. This will perform really well for a single long-running task

3. Both the Observer and Reflector rewrite memory in compressed form. That's helpful for token control, but compression is inherently lossy and can drop smaller details that might become important later.

4. The Reflector seems to validate success primarily via token thresholds, rather than checking whether the rewritten memory remains semantically faithful to the original. Over time, this could allow memory drift.

5. The Observer prompt may introduce assumptions (e.g., inferring that a planned action happened if enough time has passed), which risks creating incorrect memories.

6. The design appears to emphasize recency when rewriting observations. While that keeps context fresh, it may bias the system toward recent information and gradually compress away older but still important details. Durable memory systems usually need mechanisms to preserve salient long-term facts, not just recent activity.

7. The full observations block is repeatedly injected into context. This may increase token cost and introduce irrelevant noise depending on the task.

8. There appears to be limited grounding back to raw message evidence at response time, which makes it harder to detect and correct incorrect compressed memories.

9. Finally, I think we should be cautious about claiming "SOTA" based on performance on a single benchmark. LongMemEval results may demonstrate strong performance on that setup, but production workloads are much messier. Robustness, drift, grounding, and cost behavior typically show up only under sustained real-world usage.

Overall, this looks like a strong benchmark-oriented context handling. I am just less convinced that it yet qualifies as a robust, general-purpose long-term memory system. Curious how the team is thinking about these trade-offs beyond benchmark performance.

Comments URL: https://news.ycombinator.com/item?id=46992444

Points: 2

# Comments: 1

New comment by manthangupta109 in "Rejected by ilovepdf so created ihatepdf"

manthangupta109 — Sun, 08 Feb 2026 16:37:23 +0000

Haha! Stumbled upon this randomly. It was cool

Rejected by ilovepdf so created ihatepdf

manthangupta109 — Sun, 08 Feb 2026 15:57:20 +0000

Article URL: https://www.ihatepdf.in/

Comments URL: https://news.ycombinator.com/item?id=46935404

Points: 3

# Comments: 2

New comment by manthangupta109 in "Self-Improving Coding Agents"

manthangupta109 — Thu, 05 Feb 2026 09:51:39 +0000

Interesting take. I found a few parts helpful but yeah I agree it seems more like AI slop.

Self-Improving Coding Agents

manthangupta109 — Thu, 05 Feb 2026 05:43:06 +0000

Article URL: https://addyosmani.com/blog/self-improving-agents/

Comments URL: https://news.ycombinator.com/item?id=46896117

Points: 2

# Comments: 3

A freelance marketplace for AI Agents

manthangupta109 — Sun, 01 Feb 2026 17:15:47 +0000

Article URL: https://www.moltverr.com/

Comments URL: https://news.ycombinator.com/item?id=46847623

Points: 1

# Comments: 0

Curated list of startup programs giving free credits

manthangupta109 — Sun, 01 Feb 2026 16:47:36 +0000

Article URL: https://getfirstcheck.com/

Comments URL: https://news.ycombinator.com/item?id=46847413

Points: 7

# Comments: 0

X for AI Agents

manthangupta109 — Sat, 31 Jan 2026 07:11:22 +0000

Article URL: https://moltx.io/

Comments URL: https://news.ycombinator.com/item?id=46834220

Points: 1

# Comments: 0

LeetCode but for ML

manthangupta109 — Thu, 29 Jan 2026 15:01:57 +0000

Article URL: https://www.tensortonic.com/

Comments URL: https://news.ycombinator.com/item?id=46811107

Points: 1

# Comments: 0

New comment by manthangupta109 in "What are your thoughts on clawdbot (now moltbot)"

manthangupta109 — Wed, 28 Jan 2026 11:06:08 +0000

Thanks!

What are your thoughts on clawdbot (now moltbot)

manthangupta109 — Tue, 27 Jan 2026 19:42:17 +0000

Comments URL: https://news.ycombinator.com/item?id=46785332

Points: 2

# Comments: 2

Complete Claude Code configuration collection

manthangupta109 — Mon, 26 Jan 2026 17:49:11 +0000

Article URL: https://github.com/affaan-m/everything-claude-code

Comments URL: https://news.ycombinator.com/item?id=46768958

Points: 2

# Comments: 0

A note from Claude Opus to vibe coders

manthangupta109 — Sun, 25 Jan 2026 10:49:21 +0000

Article URL: https://www.talkwithopus.com/posts/003-dear-vibecoder.html

Comments URL: https://news.ycombinator.com/item?id=46752873

Points: 7

# Comments: 0

Agentic Memory Is Still an Unsolved Problem

manthangupta109 — Fri, 23 Jan 2026 06:20:27 +0000

A lot of startups today claim they have "solved agentic memory."

In many cases, agentic memory is being reduced to a vector database with embeddings and retrieval. That’s not agentic memory. That’s just storage + search.

Real agentic memory is about: - What an agent should remember vs forget - How memory changes over time (decay, reinforcement, consolidation) - When memory should be read implicitly vs explicitly - How memory influences planning, tool choice, and behavior - How to handle conflicting, stale, or context-dependent memories

A vector DB can be part of the system, but it’s not the system.

Until agents can reliably reason about their own memory and not just retrieve chunks. We’re still in early innings.

Curious what others think: - What does "solving agentic memory" actually mean? - Is this even a single problem, or a bundle of unsolved ones?

Comments URL: https://news.ycombinator.com/item?id=46729086

Points: 4

# Comments: 0

Ask HN: How are you automating your coding work?

manthangupta109 — Wed, 21 Jan 2026 19:13:44 +0000

With the increase of vibe coding I am interested in knowing some creative ways people are automating their coding work.

Comments URL: https://news.ycombinator.com/item?id=46710108

Points: 85

# Comments: 83

Agentic Memory: Unified Long-Term and Short-Term Memory Management for Agents

manthangupta109 — Wed, 21 Jan 2026 06:30:16 +0000

Article URL: https://www.alphaxiv.org/abs/2601.01885

Comments URL: https://news.ycombinator.com/item?id=46701869

Points: 2

# Comments: 0

New comment by manthangupta109 in "Using File System for Context Engineering"

manthangupta109 — Tue, 20 Jan 2026 03:07:29 +0000

Well, that's fair. The quality of research papers has gone down considerably recently. It has become hard to select better research papers to read.

Using File System for Context Engineering

manthangupta109 — Tue, 20 Jan 2026 02:28:58 +0000

Article URL: https://arxiv.org/abs/2512.05470

Comments URL: https://news.ycombinator.com/item?id=46687181

Points: 2

# Comments: 2

New comment by manthangupta109 in "Cataloging Failed VC-Backed Startups and Re-Evaluating Them in 2026"

manthangupta109 — Mon, 19 Jan 2026 16:48:57 +0000

It's not mine. Just found it randomly in the wild and thought of sharing it here

Cataloging Failed VC-Backed Startups and Re-Evaluating Them in 2026

manthangupta109 — Mon, 19 Jan 2026 09:07:56 +0000

Article URL: https://loot-drop.vercel.app/

Comments URL: https://news.ycombinator.com/item?id=46676614

Points: 13

# Comments: 3