AJERIS
TourBlogDocsLog inText us
Log inText us
← AI agents

Daily reflection: the 2-3 sentences that make the agent yours

Each evening the agent writes a paragraph about your day. The agent uses it to stay grounded. What it contains and why it matters more than you would expect.

Ajeris team·May 6, 2026·5 min read
A hand writing in a notebook on a wooden desk. The end-of-day reflection, two or three sentences that make the agent yours.

Each evening the agent writes a 2-3 sentence reflection about the day. Most users never read it. The agent uses it the next morning to stay grounded; the user feels the result without ever opening the manage page. This post is what the reflection contains, why it earns its keep despite being mostly invisible, and what changes when you start reading it occasionally.

The technical companion to this is the dreaming post. This one is the user-facing version. It is a small feature that punches above its weight in making the agent feel attentive.

★What is the daily reflection?

A 2-3 sentence summary the agent writes each evening, capturing the day's notable events, the threads worth carrying forward, and any skills the agent learned from solving something interesting. The agent writes it; the agent reads it the next morning. The user can read it any time.

What a real reflection looks like

A representative one, after a normal weekday for a founder.

"Yesterday Sam shipped the H2 OKR draft and got Dana's first round of feedback (push-back on the velocity number; you said you'd circle back Monday with a redrafted definition). The launch traffic looked steady through the night. You also mentioned wanting to write three pages today and didn't get there; you went to dinner with K instead, which you noted positively. You learned a small skill: when I ask for an OKR document review, lead with the controversial number, not the easy ones."

Three sentences plus one skill-learn note. The whole text is ~85 words. The agent is going to load this paragraph (plus the rest of the week's reflections) at the start of tomorrow's first conversation, which is why the morning brief feels uncannily current.

Why the paragraph matters

Three reasons you would not guess from the simple description.

Continuity. A long-running agent has a lot of episodic memory ("you said this on Tuesday at 3:14pm"). What it does not have natively is a sense of arc: "this week is shaping up around the OKR doc and the launch." The reflection compresses the arc into a paragraph, which is what loads efficiently into context the next day.

Tone calibration. The agent reads its own reflections to recalibrate. If yesterday's reflection notes "you were stressed about the launch all afternoon," the agent's morning brief tomorrow leads with the launch update first, not the inbox triage. The reflection is the input that shapes the next day's voice.

Skill capture. When the agent observes you solve something complex, it writes a candidate skill at reflection time. ("Skill: when I ask for a vendor decision, give me a 2x2 matrix instead of a list.") Some of these get promoted to the persistent skill catalog after a few repeats; others are kept ephemeral. The reflection is where this capture happens.

A pink table lamp in a calm domestic setting, the warmth of an evening pause.
The evening pause used to be a journal. Now it is a paragraph the agent writes about you.

When you should read your reflection

Most users never do. That is fine; the value is in the agent reading it. But there are a few moments when reading yours is genuinely useful.

The end of a hard week. The agent's reflection across five days is a reasonably honest narrative of what happened. Reading it can be helpful for a Sunday evening review.

When something feels off. If the morning brief has been weird ("why does the agent keep mentioning the H2 OKR doc?"), the reflection is the source of the agent's framing. Reading it reveals what the agent has been carrying forward.

A retrospective conversation. When you are trying to remember when something happened ("when did I first mention the new hire?"), the reflections are searchable: "show me the reflection that first mentions Karim." The agent surfaces the right night.

The reflections are not a journal in the human-introspection sense. They are an artifact of the agent's memory layer. Reading them is more like reading a colleague's status updates than reading your own diary.

What the agent does with skill captures

A representative skill capture from a reflection.

The agent observed that you respond to drafts of OKR documents by asking the same three clarifying questions:

  1. What's the success metric in absolute terms, not just percentage?
  2. Who owns the dependent line item?
  3. Is this on the roadmap or is it a stretch?

After the third occurrence in two weeks, the reflection notes a candidate skill: "When the user is reviewing OKR drafts, surface these three clarifying questions automatically."

The next time you ask for an OKR review, the agent uses the skill: it reads the doc and surfaces the three questions before you ask. You did not have to teach it; the agent watched, captured, and applied.

What does NOT go in the reflection

A few honest disclosures.

Sensitive content the user has not asked be remembered. If you discussed something private and asked the agent to forget, the reflection respects that.

Trivial tool calls. The reflection does not list every email send or every Spotify play. It captures themes, not transactions.

The agent's own internal reasoning. The reflection is about the user's day, not about the agent's reasoning loop. The reasoning is in the audit trail; the reflection is the user-facing summary.

How to tune your reflections

If the reflections feel off (too detailed, too brief, missing categories you care about), you tune them in plain English.

You: my evening reflections are too focused on work. include
     personal stuff too - what i ate, who i saw, the mood.
 
Ajeris: Got it. The reflection will broaden. Tonight's will
        include personal threads alongside work.
 
[Next evening]
> "Today you shipped the OKR draft revision and felt good about
   it. Lunch with K at the place on Highland (you noted the
   wait was long). Long evening walk before dinner. You
   mentioned wanting to write more this week and didn't get
   to it; that's been the same pattern for ten days."

Tuning takes one sentence. The next reflection picks up the change.

  1. 1
    Evening
    Reflection runs per-user.
  2. 2
    85-150 words
    Captures themes, threads, mood.
  3. 3
    Skill capture
    Notes candidate skills if observed.
  4. 4
    Loaded next day
    Used as context for the morning brief.
  5. 5
    User-readable
    Search and read on demand.
Without a daily reflection
  • Agent has episodic memory but no narrative arc.
  • Morning brief feels generic, recency-biased.
  • Skills captured manually if at all.
  • No clean way to ask 'what was I working on last Tuesday.'
With Ajeris
  • Agent carries the week's narrative forward.
  • Morning brief reflects the actual arc.
  • Skills captured automatically; you accept or reject.
  • Reflections are searchable. The week has a record.
eveningReflection time
~100 wordsAverage length
autoSkill capture

Frequently asked

Can I read past reflections?+
Yes. 'Show me last Tuesday's reflection.' 'Show me reflections from August.' The agent surfaces what you ask for. They are also searchable by content.
Can I edit a reflection if it gets something wrong?+
Yes. 'The reflection has X wrong; the actual thing was Y.' The agent updates and notes the correction. The original is preserved with a pointer to the corrected version.
What if I want my reflections written in the morning instead of evening?+
Tell the agent. The schedule is configurable. Some users prefer morning reflections that look back at yesterday; the default is evening because it captures the day fresh.
Can I disable reflections?+
Yes, but the morning brief quality drops. Reflections are the connective tissue between days; without them the agent loses a lot of its sense of arc across days. We do not recommend disabling but you can.
Are reflections shared with anyone?+
Per-user, never shared. The reflection is yours. The agent uses it; nobody else sees it.

Try it

Read tonight's reflection. The first time you see your own week summarized in a paragraph the agent wrote, you understand why it matters more than the simple description suggests.

More stories

A phone on a cafe table, top-down. The case for the agent that lives where the conversation already is.
AI agents

Why texting wins: the case for messaging-first agents

Chat windows, smart speakers, and dedicated apps each have a thesis for the AI agent. The thesis that holds up daily is the one that meets people in their messages.

An iPhone on a cafe table with a serif headline reading 'An AI agent that lives in your text messages.'
AI agents

What is a personal AI agent?

A personal AI agent is one assistant that lives in your text messages and actually does the work across every tool you already use. Here is what that means.

Try Ajeris.

Text +1 (470) 821 4265. Your agent replies in seconds. Free.

Text on WhatsApp
Connects to

Every tool you already use. Your agent does the rest.

WhatsApp
Gmail
Calendar
Slack
Notion
GitHub
Instagram
Spotify
Uber
DoorDash
Apple Music
Telegram
WhatsApp
Gmail
Calendar
Slack
Notion
GitHub
Instagram
Spotify
Uber
DoorDash
Apple Music
Telegram
Shopify
QuickBooks
Xero
Square
HubSpot
Salesforce
Stripe
Lyft
Alexa
Google LSA
Hue
Claude
Shopify
QuickBooks
Xero
Square
HubSpot
Salesforce
Stripe
Lyft
Alexa
Google LSA
Hue
Claude

Message to start

+1 470 821 4265

Message the number above on WhatsApp or by SMS. Your assistant replies within seconds and sets itself up in the conversation. Or sign up by web form to be added to our list.

Welcome message. “Hi. I’m Ajeris. I’ll be your personal agent. What should I call you? Reply STOP to unsubscribe, HELP for help.”

STOP to unsubscribe. HELP for support.

Message and data rates may apply. Carriers are not liable for delayed or undelivered messages. Message frequency varies based on your interaction with the agent.

Questions: hi@ajeris.com

Your assistant is already in your phone. You just need to text it.

The personal assistant you can actually get today. No app, no waitlist.

WhatsApphi@ajeris.com
Personal
  • Channels
  • Morning brief
  • Meetings
  • Research and build
  • Laptop on a leash
  • Everyday
  • Memory
Platform
  • Connectors
  • Platform
Developers
  • Docs
  • Quickstart
  • Platform overview
  • Memory model
  • Skills
  • Scheduled tasks
  • MCP server
  • Claude Code
  • Alexa+
Resources
  • Blog
  • Compare
  • Glossary
  • Live webinar
  • Creator Program
Trust
  • Privacy
  • Security
  • Terms
  • Data retention
  • Access controls
  • Vulnerability management
  • Opt in
  • Disconnect

Your personal agent, by text.

Try it free·Ajeris for business·docs.ajeris.com·© 2026 Ajeris